Practical Applications of Computational Biology and Bioinformatics 16th International Conference PACBB 2022 Florentino Fdez-Riverola

Full download test bank at ebookmeta.
com
Practical Applications of Computational Biology

and Bioinformatics 16th International Conference
PACBB 2022 Florentino Fdez-Riverola
For dowload this book click LINK or Button below
https://ebookmeta.com/product/practical-
applications-of-computational-biology-and-
bioinformatics-16th-international-conference-
pacbb-2022-florentino-fdez-riverola/
OR CLICK BUTTON
DOWLOAD EBOOK
Download More ebooks from https://ebookmeta.com

More products digital (pdf, epub, mobi) instant
download maybe you interests ...
GeNeDis 2022 Computational Biology and Bioinformatics

Advances in Experimental Medicine and Biology 1424
Panagiotis Vlamos (Editor)
https://ebookmeta.com/product/genedis-2022-computational-biology-
and-bioinformatics-advances-in-experimental-medicine-and-
biology-1424-panagiotis-vlamos-editor/
Bioinformatics and Computational Biology 1st Edition

Hamid R. Arabnia
https://ebookmeta.com/product/bioinformatics-and-computational-
biology-1st-edition-hamid-r-arabnia/
Parallel Computational Technologies 16th International

Conference PCT 2022 Dubna Russia March 29 31 2022
Revised Selected Papers Communications in Computer and
Information Science 1618 Leonid Sokolinsky
https://ebookmeta.com/product/parallel-computational-
technologies-16th-international-conference-pct-2022-dubna-russia-
march-29-31-2022-revised-selected-papers-communications-in-
computer-and-information-science-1618-leonid-sokolin/
Bioinformatics and Computational Biology A Primer for

Biologists 1st Edition Basant K. Tiwary
https://ebookmeta.com/product/bioinformatics-and-computational-
biology-a-primer-for-biologists-1st-edition-basant-k-tiwary/
Advances in Practical Applications of Agents Multi
Agent Systems and Complex Systems Simulation The PAAMS
Collection 20th International Conference PAAMS 2022
Frank Dignum
https://ebookmeta.com/product/advances-in-practical-applications-
of-agents-multi-agent-systems-and-complex-systems-simulation-the-
paams-collection-20th-international-conference-paams-2022-frank-
dignum/
Bioinformatics: A Practical Guide to Next Generation

Sequencing Data Analysis (Chapman & Hall/CRC
Computational Biology Series) 1st Edition Hamid D.
Ismail
https://ebookmeta.com/product/bioinformatics-a-practical-guide-
to-next-generation-sequencing-data-analysis-chapman-hall-crc-
computational-biology-series-1st-edition-hamid-d-ismail/
Enterprise, Business-Process and Information Systems

Modeling: 23rd International Conference, BPMDS 2022 and
27th International Conference, EMMSAD 2022 Adriano
Augusto
https://ebookmeta.com/product/enterprise-business-process-and-
information-systems-modeling-23rd-international-conference-
bpmds-2022-and-27th-international-conference-emmsad-2022-adriano-
augusto/
Algorithmic Aspects in Information and Management: 16th

International Conference, AAIM 2022, Guangzhou, China,
August 13–14, 2022, Proceedings (Lecture Notes in
Computer Science, 13513) Qiufen Ni
https://ebookmeta.com/product/algorithmic-aspects-in-information-
and-management-16th-international-conference-aaim-2022-guangzhou-
china-august-13-14-2022-proceedings-lecture-notes-in-computer-
science-13513-qiufen/
Advances on Broad Band Wireless Computing Communication

and Applications Proceedings of the 16th International
Conference on Broad Band Wireless 1st Edition Leonard
Barolli
https://ebookmeta.com/product/advances-on-broad-band-wireless-
computing-communication-and-applications-proceedings-of-the-16th-
international-conference-on-broad-band-wireless-1st-edition-
Lecture Notes in Networks and Systems 553
Florentino Fdez-Riverola ·
Miguel Rocha ·
Mohd Saberi Mohamad ·
Simona Caraiman ·
Ana Belén Gil-González Editors
Practical Applications
of Computational
Biology and
Bioinformatics,
16th International
Conference
(PACBB 2022)
Lecture Notes in Networks and Systems
Volume 553
Series Editor
Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences,
Warsaw, Poland
Advisory Editors
Fernando Gomide, Department of Computer Engineering and Automation—DCA,
School of Electrical and Computer Engineering—FEEC, University of
Campinas—UNICAMP, São Paulo, Brazil
Okyay Kaynak, Department of Electrical and Electronic Engineering,
Bogazici University, Istanbul, Turkey
Derong Liu, Department of Electrical and Computer Engineering, University of
Illinois at Chicago, Chicago, USA
Institute of Automation, Chinese Academy of Sciences, Beijing, China
Witold Pedrycz, Department of Electrical and Computer Engineering, University of
Alberta, Alberta, Canada
Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland
Marios M. Polycarpou, Department of Electrical and Computer Engineering,
KIOS Research Center for Intelligent Systems and Networks, University of Cyprus,
Nicosia, Cyprus
Imre J. Rudas, Óbuda University, Budapest, Hungary
Jun Wang, Department of Computer Science, City University of Hong Kong,
Kowloon, Hong Kong
The series “Lecture Notes in Networks and Systems” publishes the latest
developments in Networks and Systems—quickly, informally and with high quality.
Original research reported in proceedings and post-proceedings represents the core
of LNNS.
Volumes published in LNNS embrace all aspects and subfields of, as well as new
challenges in, Networks and Systems.
The series contains proceedings and edited volumes in systems and networks,
spanning the areas of Cyber-Physical Systems, Autonomous Systems, Sensor
Networks, Control Systems, Energy Systems, Automotive Systems, Biological
Systems, Vehicular Networking and Connected Vehicles, Aerospace Systems,
Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power Systems,
Robotics, Social Systems, Economic Systems and other. Of particular value to
both the contributors and the readership are the short publication timeframe and
the world-wide distribution and exposure which enable both a wide and rapid
dissemination of research output.
The series covers the theory, applications, and perspectives on the state of the art
and future developments relevant to systems and networks, decision making, control,
complex processes and related areas, as embedded in the fields of interdisciplinary
and applied sciences, engineering, computer science, physics, economics, social, and
life sciences, as well as the paradigms and methodologies behind them.
Indexed by SCOPUS, INSPEC, WTI Frankfurt eG, zbMATH, SCImago.
All books published in the series are submitted for consideration in Web of Science.
For proposals from Asia please contact Aninda Bose (aninda.bose@springer.com).

Florentino Fdez-Riverola · Miguel Rocha ·
Mohd Saberi Mohamad · Simona Caraiman ·
Ana Belén Gil-González
Editors
Practical Applications
of Computational Biology
and Bioinformatics, 16th
International Conference
(PACBB 2022)
Editors
Florentino Fdez-Riverola Miguel Rocha
Computer Science Department Campus de Gualtar
Universidad de Vigo Universidade do Minho
Vigo, Spain Braga, Portugal
Mohd Saberi Mohamad Simona Caraiman

College of Medicine and Health Sciences Gheorghe Asachi Technical University
United Arab Emirates University of Ias, i
Al Ain, Abu Dhabi, United Arab Emirates Ias, i, Romania
Ana Belén Gil-González

Edificio I+D+i
University of Salamanca
Salamanca, Spain
ISSN 2367-3370 ISSN 2367-3389 (electronic)

Lecture Notes in Networks and Systems
ISBN 978-3-031-17023-2 ISBN 978-3-031-17024-9 (eBook)
https://doi.org/10.1007/978-3-031-17024-9
© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
The success of bioinformatics in recent years has been prompted by research in

molecular biology and molecular medicine in several initiatives. These initiatives
gave rise to an exponential increase in the volume and diversification of data,
including nucleotide and protein sequences and annotations, high-throughput exper-
imental data, biomedical literature, among many others. Systems biology is a related
research area that has been replacing the reductionist view that dominated biology
research in the last decades, requiring the coordinated efforts of biological researchers
with those related to data analysis, mathematical modelling, computer simulation and
optimization.
The accumulation and exploitation of large-scale databases prompt new compu-
tational technology and for research into these issues. In this context, many widely
successful computational models and tools used by biologists in these initiatives,
such as clustering and classification methods for gene expression data, are based on
computer science/artificial intelligence (CS/AI) techniques. In fact, these methods
have been helping in tasks related to knowledge discovery, modelling and optimiza-
tion tasks, aiming at the development of computational models so that the response of
biological complex systems to any perturbation can be predicted. The 16th Interna-
tional Conference on Practical Applications of Computational Biology & Bioinfor-
matics (PACBB) aims to promote the interaction among the scientific community to
discuss applications of CS/AI with an interdisciplinary character, exploring the inter-
actions between sub-areas of CS/AI, bioinformatics, chemoinformatic and systems
biology. The PACBB’22 technical programme includes ten papers of authors from
many different countries (Bahrain, Canada, France, Italy, Portugal, Saudi Arabia,
Spain and UK) and different subfields in bioinformatics and computational biology.
All papers underwent a peer review selection: each paper was assessed by three
different reviewers from an international panel composed of about 46 members from
11 countries. The quality of submissions was on average good, with an acceptance
rate of approximately 60% (10 accepted papers from 15 submissions).
There will be special issues in JCR-ranked journals, such as Interdisciplinary
sciences: mathematical biosciences and engineering, integrative bioinformatics,
information fusion, neurocomputing, sensors, processes and electronics. Therefore,
v
vi Preface
this event will strongly promote the interaction among researchers from international
research groups working in diverse fields. The scientific content will be innovative,
and it will help improve the valuable work that is being carried out by the participants.
This symposium is organized by the University of L’Aquila (Italy) with the
collaboration of the United Arab Emirates University, the University of Minho, the
University of Vigo, the University of Salamanca and the Gheorghe Asachi Technical
University of Ias, i. We would like to thank all the contributing authors, the members
of the programme committee and the sponsors. We thank for funding support to
the project: “Intelligent and sustainable mobility supported by multi-agent systems
and edge computing” (Id. RTI2018-095390-B-C32), and finally, we thank the local
organization members for their valuable work, which is essential for the success of
PACBB’22.
Vigo, Spain Florentino Fdez-Riverola

Braga, Portugal Miguel Rocha
Al Ain, United Arab Emirates Mohd Saberi Mohamad
Ias, i, Romania Simona Caraiman
Salamanca, Spain Ana Belén Gil-González
Acknowledgements
vii
Organization
Program Committee Chairs
Mohd Saberi Mohamad, United Arab Emirates University, United Arab Emirates
Miguel Rocha, University of Minho, Portugal
Organising Committee Chairs
Florentino Fdez-Riverola, University of Vigo, Spain

Ana Belén Gil-González, University of Salamanca, Spain
Simona Caraiman, Gheorghe Asachi Technical University of Ias, i, Romania
Advisory Committee
Grabriella Panuccio, Istituto Italiano di Tecnología, Italy
Local Organizing Committee
Pierpaolo Vittorini (Co-chair), University of L’aquila, Italy

Tania Di Mascio (Co-chair), University of L’aquila, Italy
Federica Caruso, University of L’Aquila, Italy
Anna Maria Angelone, University of L’Aquila, Italy
ix
x Organization
Organizing Committee
Juan M. Corchado Rodríguez, University of Salamanca, Spain; AIR Institute, Spain

Fernando De la Prieta, University of Salamanca, Spain
Sara Rodríguez González, University of Salamanca, Spain
Javier Prieto Tejedor, University of Salamanca, Spain; AIR Institute, Spain
Pablo Chamoso Santos, University of Salamanca, Spain
Liliana Durón, University of Salamanca, Spain
Belén Pérez Lancho, University of Salamanca, Spain
Ana Belén Gil González, University of Salamanca, Spain
Ana De Luis Reboredo, University of Salamanca, Spain
Angélica González Arrieta, University of Salamanca, Spain
Emilio S. Corchado Rodríguez, University of Salamanca, Spain
Alfonso González Briones, University of Salamanca, Spain
Yeray Mezquita Martín, University of Salamanca, Spain
Beatriz Bellido, University of Salamanca, Spain
María Alonso, University of Salamanca, Spain
Sergio Marquez, University of Salamanca, Spain
Marta Plaza Hernández, University of Salamanca, Spain
Guillermo Hernández González, AIR Institute, Spain
Ricardo S. Alonso Rincón, University of Salamanca, Spain
Raúl López, University of Salamanca, Spain
Sergio Alonso, University of Salamanca, Spain
Andrea Gil, University of Salamanca, Spain
Javier Parra, University of Salamanca, Spain
Programme Committee
Vera Afreixo, University of Aveiro, Portugal

Manuel Álvarez Díaz, University of A Coruña, Spain
Joel P. Arrais, University of Coimbra, Portugal
Carlos Bastos, University of Aveiro, Portugal
Lourdes Borrajo, University of Vigo, Spain
Ana Cristina Braga, University of Minho, Portugal
Rui Camacho, University of Porto, Portugal
Angel Canal, Universidad de Salamanca, Spain
Yingbo Cui, National University of Defense Technology, China
Sergio Deusdado, IPB-Polytechnic Institute of Bragança, Portugal
Oscar Dias, University of Minho, Portugal
Nuno Filipe, University of Porto, Portugal
Dino Franklin, Federal University of Uberlandia, Brazil
Narmer Galeano, Universidad Catolica de Manizales, Colombia
Organization xi
Rosalba Giugno, University of Verona, Italy

Gustavo Isaza, University of Caldas, Colombia
Paula Jorge, IBB - CEB Centre of Biological Engineering, Portugal
Rosalia Laza, Universidad de Vigo, Spain
Thierry Lecroq, University of Rouen, France
Giovani Librelotto, Universidade Federal de Santa Maria, Brazil
Filipe Liu, Data Science and Learning Division, Argonne National Laboratory, USA
Hugo López Fernández, Instituto de Investigação e Inovação em Saúde (i3S), Spain
Eva Lorenzo Iglesias, University of Vigo, Spain
Gonçalo Marques, Polytechnic of Coimbra, Portugal
Mohd Saberi Mohamad, United Arab Emirates University, United Arab Emirates
Loris Nanni, University of Padua, Italy
José Luis Oliveira, University of Aveiro, Portugal
Vitor Pereira, University of Minho, Portugal
Armando Pinho, University of Aveiro, Portugal
Ignacio Ponzoni, CONICET, Argentina
Miguel Reboiro-Jato, University of Vigo, Spain
Jose Ignacio Requeno, University of Zaragoza, Spain
João Manuel Rodrigues, DETI/IEETA, University of Aveiro, Portugal
Gustavo Santos García, Universidad de Salamanca, Spain
Ana Margarida Sousa, University of Minho, Portugal
Niclas Ståhl, University of Skövde, Sweden
Carolyn Talcott, SRI International, USA
Rita Margarida Teixeira Ascenso, ESTG–IPL, Portugal
Antonio J. Tomeu-Hardasmal, University of Cadiz, Spain
Eduardo Valente, IPCB, Portugal
Alejandro F. Villaverde, Instituto de Investigaciones Marinas (CSIC), Spain
Pierpaolo Vittorini, University of L’Aquila, Italy
Contents
TooT-BERT-T: A BERT Approach on Discriminating Transport

Proteins from Non-transport Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Hamed Ghazikhani and Gregory Butler
Machine Learning and Deep Learning Techniques for Epileptic
Seizures Prediction: A Brief Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Marco Hernández, Ángel Canal-Alonso, Fernando de la Prieta,
Sara Rodríguez, Javier Prieto, and Juan Manuel Corchado
The Covid-19 Decision Support System (C19DSS) – A Mobile App . . . . . 23
Pierpaolo Vittorini, Nicolò Casano, Gaia Sinatti,
Silvano Junior Santini, and Clara Balsano
Towards a Flexible and Portable Workflow for Analyzing
miRNA-Seq Neuropsychiatric Data: An Initial Replicability
Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Daniel Pérez-Rodríguez, Mateo Pérez-Rodríguez,
Roberto C. Agís-Balboa, and Hugo López-Fernández
The NAD Interactome, Identification of Putative New
NAD-Binding Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Sara Duarte-Pereira, Sérgio Matos, José Luís Oliveira,
and Raquel M. Silva
Multiple Instance Learning Based on Mol2vec Molecular
Substructure Embeddings for Discovery of NDM-1 Inhibitors . . . . . . . . . 55
Thomas Papastergiou, Jérôme Azé, Sandra Bringay, Maxime Louet,
Pascal Poncelet, and Laurent Gavara
Towards Improving Bio-Image Segmentation Quality Through
Ensemble Post-processing of Deep Learning and Classical 3D
Segmentation Pipelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Anuradha Kar
xiii
xiv Contents
Exploring Xylella fastidiosa’s Metabolic Traits Using a GSM

Model of the Phytopathogenic Bacterium . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Alexandre Oliveira, Emanuel Cunha, Miguel Silva, Cristiana Faria,
and Oscar Dias
Genomic Regions with Atypical Concentration of Inverted Repeats . . . . 89
Carlos A. C. Bastos, Vera Afreixo, João M. O. S. Rodrigues,
and Armando J. Pinho
EvoPPI 2: A Web and Local Platform for the Comparison
of Protein–Protein Interaction Data from Multiple Sources
from the Same and Distinct Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Miguel Reboiro-Jato, Jorge Vieira, Sara Rocha, André D. Sousa,
Hugo López-Fernández, and Cristina P. Vieira
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

TooT-BERT-T: A BERT Approach
on Discriminating Transport Proteins
from Non-transport Proteins
Hamed Ghazikhani and Gregory Butler
Abstract Transmembrane transport proteins (transporters) serve a crucial role for

the transport of hydrophilic molecules across hydrophobic membranes in every living
cell. The structures and functions of many membrane proteins are unknown due to the
enormous effort required to characterize them. This article proposes TooT-BERT-T, a
technique that employs the BERT representation to analyze and discriminate between
transporters and non-transporters using a Logistic Regression classifier. Additionally,
we evaluate frozen and fine-tuned representations from two different BERT models.
Compared to state-of-the-art prediction methods, TooT-BERT-T achieves the highest
accuracy of 93.89% and MCC of 0.86.
Keywords Transmembrane transport proteins · Machine learning · BERT ·

Language model · Transformers · Neural network
1 Introduction
Around one-third of the proteins in a cell are found in its membrane, and approxi-
mately one-third of these proteins are involved in molecule transport [21]. Trans-
membrane transport proteins, also known as transporters, are required for cell
metabolism, ion homeostasis, signal transduction, binding with small molecules in
the extracellular space, immune recognition, energy transduction, and physiological
and developmental processes [21].
Protein research has advanced our knowledge of human health and disease treat-
ment. The decreasing cost of sequencing technology has enabled the generation of
H. Ghazikhani (B) · G. Butler

Department of Computer Science and Software Engineering, Concordia University,
Montreal, Canada
e-mail: hamed.ghazikhani@concordia.ca
G. Butler
e-mail: gregory.butler@concordia.ca
G. Butler
Centre for Structural and Functional Genomics, Concordia University, Montreal, Canada
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 1
F. Fdez-Riverola et al. (eds.), Practical Applications of Computational Biology
and Bioinformatics, 16th International Conference (PACBB 2022), Lecture Notes
in Networks and Systems 553, https://doi.org/10.1007/978-3-031-17024-9_1
2 H. Ghazikhani and G. Butler
massive datasets of naturally occurring proteins with enough information to build

sophisticated machine learning models of protein sequences [23].
Since proteins, like human languages, are denoted by string concatenation, we
can apply natural language processing (NLP) approaches [18]. Transformer neural
networks (Transformers) have contributed significantly to the field of natural lan-
guage processing [22]. Autoencoders, for example, BERT (Bidirectional Encoder
Representations from Transformers) [9], are stacking models that are trained by cor-
rupting input tokens and attempting to recover the original sentence [11]. While they
can generate text as well, they are typically used to create vector representations for
future tasks such as classification [11].
A massive collection of protein sequences from UniProt Archive (UniParc) [14]
and the Big Fantastic Database (BFD) [11, 13] comprising over 390 billion amino
acids resulted in ProtTrans [10], an amazing adaption to the protein domain of six
available Transformer topologies which are Transformer-XL, BERT, Albert, XLnet,
T5, and Electra.
TooT-BERT-T proposes a method for discriminating transport proteins from non-
transport proteins using representations from ProtBERT-BFD and Logistic Regres-
sion. Our investigation can be summarised as follows: 1) Using ProtBERT-BFD
to discriminate between transport and non-transport proteins for the first time. 2)
Evaluation of frozen/fine-tuned ProtBERT-BFD representations. 3) Evaluation of
frozen/fine-tuned MembraneBERT representations. 4) The fine-tuned Transporter-
BERT is a publicly accessible model pre-trained on the BFD database and fine-
tuned using the transport proteins dataset (https://huggingface.co/ghazikhanihamed/
TransporterBERT). 5) Proposing TooT-BERT-T as a method for classifying transport
proteins that outperforms all other approaches.
The following is the outline for the paper: Sect. 2 describes the related work.
Section 3 contains information about the dataset and experimental design used in this
study. Section 4 compares and analyses the outcomes of TooT-BERT-T and Sect. 5
brings the paper to a close.
2 Related Work
Aplop and Butler [4, 5] provide a comprehensive overview of transport protein

prediction methods. Earlier efforts used experimentally characterized databases to
conduct homology searches for novel transporters. For example, TransATH [5] auto-
mates the Saier’s protocol via sequence similarity. TransATH improves transmem-
brane segment computations by including subcellular localization and claims an
overall accuracy of 71.0%.
TrSSP (Transporter Substrate Specificity Prediction Service) [16] was developed
to predict the substrate category of membrane transport proteins in an attempt to
overcome the limitations of homology methods. The TrSSP tool predicts top-level
transporters with an accuracy of 78.99 and 80.00%, respectively, and an MCC of
0.58 and 0.57 on the cross-validation and independent test sets.
Transporters Prediction Using BERT 3
SCMMTP [15] makes use of a novel scoring card method (SCM) to ascertain the
dipeptide composition of potential membrane transport proteins. SCMMTP begins
with a 400-dipeptide starting matrix and scores dipeptides based on the difference
between positive and negative compositions. Following that, the matrix is optimized
using a genetic algorithm. SCMMTP achieved an overall accuracy of 81.12% and
76.11% and an MCC of 0.62 and 0.47, respectively, on the training and independent
datasets.
Nguyen et al. [17] characterize transporter protein sequences using a word-
embedding technique. The protein sequence is defined by the word embedding and
the protein’s biological terms frequency. They achieved accurate results in terms
of transporter substrate specificity but not in terms of transporter detection. When
cross-validation was used, the prediction accuracy for transporters was only 83.94
and 85.00% using the independent dataset.
In 2020, Alballa and Butler developed TooT-T [2], an ensemble technique that
combines the results of two distinct approaches: homology annotation transfer and
machine learning. BLAST searches the Transporter Classification Database (TCDB)
[20] for homology to a query protein. If a query meets three thresholds, it is pro-
jected as a transporter. It also computes three composition features for training their
respective SVM models. Finally, the meta-model assigns a protein the transport pro-
tein classification. They claim accuracy of 90.07% and 92.22%, respectively, and
MCC values of 0.80 and 0.82 for the cross-validation and independent test sets,
respectively. While incorporating multiple feature sets and classifiers improves the
classification of transport proteins in TooT-T, it also increases the task’s complexity.
3 Materials and Methods
3.1 Dataset
This work utilizes the dataset from the TrSSP project [16] which can be accessed
at the following URL: https://www.zhaolab.org/TrSSP/. The dataset was created
using the UniProt database [14], in which 10, 780 transporter, carrier, and channel
proteins were initially well characterized at the protein level with different substrate
specificity annotation. Mishra et al. [16] eliminated from this benchmarking dataset
fragmented sequences, sequences with more than two substrate specificities, and
biological function annotations based only on sequence similarity. As presented in
Table 1 the final dataset contains 1, 560 protein sequences for the training and test
sets. This dataset is referred to as DS-T, which stands for a dataset for transporter
proteins.
Table 1 DS-T: transport proteins dataset

Class Training Test Total
Transporter 780 120 900
Nontransporter 600 60 660
Total 1,380 180 1,560
3.2 Protein Sequence Representation
As multiple studies demonstrate, representation learning, a branch of machine learn-

ing in which the representation is estimated concurrently with the statistical model,
is gaining traction in biology. Works [3, 6, 19] highlight how representations can
assist in extracting crucial biological information from the millions of observations
collected by modern sequencing technologies [8].
BERT (Bidirectional Encoder Representations from Transformers) [9] is a lan-
guage model used in natural language processing that employs a multi-layer bi-
directional Transformer encoder that employs an attention mechanism in each
encoder layer to condition both left and right context and process all words in the
sentence in parallel. Each encoder layer comprises two sub-layers: multi-head self-
attention and feed-forward neural networks. While encoding a specific word, the
multi-head self-attention sublayer assists the encoder in looking at other words in
the input sentence. The following formula is used to compute the scaled dot-product
attention sublayer [22]:
Multi H ead(Q, K , V ) = Concat (head1 , ..., headn )W o (1)

headi = Attention(QWiQ , K WiK , V WiV ) (2)
T
QK
Attention(Q, K , V ) = so f tmax( √ )V (3)
dk
where Q (Query), K (Key) and V (Value) are various linear transformations of the
input features in order to obtain information representations for various subspaces.
The dimension of K is dk and WiQ , WiK , WiV and WiO are weight matrices.
BERT is a two-step framework: pre-training and fine-tuning. Pre-training is train-
ing the model on a large amount of unlabeled data in an unsupervised manner. In
contrast, fine-tuning is the process of initializing the model with the pre-trained
parameters and fine-tuning all parameters using labeled data from downstream tasks
via an additional classifier [9].
There are two methods for extracting representations from pre-trained BERT
models: (i) frozen and (ii) fine-tuned. The former extracts features from a pre-trained
BERT model without updating the model’s weights, whereas the latter extracts
features after training the pre-trained BERT model on a smaller dataset and fine-
tuning the model’s weights [9].
ProtBERT-BFD [10] is the BERT model which has been pre-trained on a large
corpus of protein sequences from the BFD database (https://bfd.mmseqs.com) which
contains 2.5 billion protein sequences. MembraneBERT is ProtBERT-BFD fine-
tuned using the TooT-M membrane proteins dataset [1]. MembraneBERT can be
found at (https://huggingface.co/ghazikhanihamed/MembraneBERT).
The representations from the final hidden layer of ProtBERT-BFD and Mem-
braneBERT models are used in conjunction with a mean-pooling strategy, which is
concluded to be the optimal method in ProtTrans [10].
3.3 Fine-Tuning a BERT Model
We add a classification layer and train the entire BERT model on the transporters
training set to fine-tune a BERT model. We randomly chose 10% of the training
samples as the validation set in this study. The downstream task dataset will update all
initialized weights from pre-training during the fine-tuning phase. We fine-tuned the
BERT models using the Trainer API from HuggingFace [24]. This is a preliminary
investigation of BERT’s role in transport protein analysis, so we used the same
hyperparameter settings as ProtTrans [10], except for the empirically determined
number of training epochs of 13 for ProtBERT-BFD and 10 for MembraneBERT.
We discovered these numbers when we have the maximum performance throughout
the validation set results. Additional hyperparameters for fine-tuning are listed in
Table 2 which are recommended and used in ProtTrans project.
Table 2 Fine-tuning ProtBERT-BFD and MembraneBERT hyperparameters

Hyperparamer Value
Training batch size 1
Evaluation batch size 32
Warmup steps 1000
Weight decay 0.01
Gradient accumulation steps 64
Except for the training epochs, ProtBERT-BFD and MembraneBERT use the same fine-tuning
hyperparameter settings as ProtTrans [10].
3.4 Logistic Regression
Logistic Regression is a widely used classification technique in medical/biological

research [12]. The Logistic Regression algorithm used was the scikit-learn Python
module (https://scikit-learn.org) and the study used the default hyperparameters.
3.5 Evaluation
A 10-fold cross-validation (CV) technique was used in this analysis to evaluate the
model’s performance by partitioning the dataset into ten sections. For the purpose of
fine-tuning the BERT, 10% of the training set was used as the validation set, while
the remaining 90% was used for training. The independent test set is utilised for the
sole purpose of evaluating the method.
3.6 Evaluation Metrics
Four key evaluation criteria are considered in this project: Sensitivity (Sen), Speci-
ficity (Spc), Accuracy (Acc), and MCC.
(T P × T N ) − (F P × F N )
MCC = √ (4)
(T P + F P)(T P + F N )(T N + F P)(T N + F N )
MCC is an acronym for Matthew’s Correlation Coefficient. For imbalanced data,

MCC is a more stable assessment metric [7].
4 Results and Discussion
4.1 Fine-Tuning ProtBERT-BFD and MembraneBERT
We compared both representations of ProtBERT-BFD and MembraneBERT, without

(frozen) and with (fine-tuned) fine-tuning using the DS-T dataset. Figure 1 visualises
the effect of fine-tuning ProtBERT-BFD and MembraneBERT for each epoch.
As demonstrated, the ProtBERT-BFD model improved representations in each
epoch, increasing from zero MCC and 56% accuracy to 0.77 MCC and 87% accuracy
on the validation set. The ProtBERT-BFD model outperforms the MembraneBERT
model, indicating that a BERT model trained on a more extensive set of protein
sequences has superior representation and performance in the downstream task
fine-tuning. Additionally, the ProtBERT-BFD performs better in both frozen and
Fig. 1 The effect of fine-tuning (This figure depicts the results of fine-tuning the ProtBERT-BFD
(left) and MembraneBERT (right) with accuracy and MCC metrics at each epoch on the validation
set. The y-axis and x-axis display the scores and epochs, respectively)
Table 3 Logistic Regression performance with ProtBERT-BFD and MembraneBERT

Model Sen (%) Spc (%) Acc (%) MCC
Ind. CV Ind. CV Ind. CV Ind. CV
ProtBERT-BFD 76.67 80.00 90.83 82.69 86.11 81.52 0.6840 0.6262
frozen
ProtBERT-BFD 95.83 96.79 90.00 97.17 93.89 96.96 0.8620 0.9387
fine-tuned
MembraneBERT 88.33 80.51 68.33 77.50 81.67 79.20 0.5799 0.5797
frozen
MembraneBERT 86.67 98.08 85.00 97.00 86.11 97.61 0.6989 0.9512
fine-tuned
This table summarizes the 10-fold CV and independent test set performance of frozen/fine-tuned
representations from the ProtBERT-BFD and MembraneBERT models in terms of sensitivity, speci-
ficity, accuracy, and MCC. The maximum value for each column is displayed in boldface.
fine-tuned representations than MembraneBERT, with the exception of the frozen

representation of sensitivity. Despite the high cost of fine-tuning the 420 million-
parameter ProtBERT-BFD model, our results (Table 3) demonstrate that fine-tuning
ProtBERT-BFD for transport protein prediction results in the best representation.
4.2 Logistic Regression with Fine-Tuned ProtBERT-BFD
We selected Logistic Regression as a preliminary good binary classifier because it

is simple to implement and interpret, has been tested in the ProtTrans project, and
produces competitive results [10]. Table 3 demonstrates that Logistic Regression with
both fine-tuned ProtBERT-BFD and MembraneBERT representations performs well,

with fine-tuned ProtBERT-BFD outperforming MembraneBERT on all independent
test set results, while MembraneBERT outperforms sensitivity, accuracy, and MCC
on CV results.
4.3 Comparison of TooT-BERT-T with State-of-the-Art

Models
Table 4 and Fig. 2 are used to compare TooT-BERT-T to other published methods that
use only the protein sequence on the same dataset. As demonstrated, TooT-BERT-T
outperforms other published works in all evaluation metrics except sensitivity, where
Nguyen et al. [17] achieves 100% sensitivity.
Table 4 Comparative performance of TooT-BERT-T with state-of-the-art

Method Sen (%) Spc (%) Acc (%) MCC
Ind. CV Ind. CV Ind. CV Ind. CV
SCMMTP [15] 80.00 83.76 68.33 77.68 76.11 81.12 0.47 0.62
TrSSP [16] 76.67 76.67 81.67 78.46 80.00 78.99 0.57 0.58
Nguyen et al. [17] 100.00 83.14 77.50 84.48 85.00 83.94 0.73 0.68
TooT-T [2] 94.17 90.15 88.33 89.97 92.22 90.07 0.82 0.80
TooT-BERT-T 95.83 96.79 90.00 97.17 93.89 96.96 0.86 0.94
This table compares the outcomes of various techniques using sensitivity, specificity, accuracy, and
MCC metrics on the CV and independent test set. Results taken from [2]. The maximum value for
each column is displayed in boldface.
Fig. 2 Comparison of methodologies

Fig. 3 TooT-BERT-T Predicted values

confusion matrix (This figure
summarises the performance
Actual values
of TooT-BERT-T, where T T 115 5
represents transport protein
and non-T represents
non-transport protein)
non-T 6 54
T non-T
TooT-BERT-T has a greater specificity (rate of true negatives) than the approach of
Nguyen et al. [17], indicating that it makes fewer false positive predictions (Fig. 3).
This is essential for achieving a high true negative rate of 90% when describing
non-transport proteins.
The proposed method, TooT-BERT-T, which employs fine-tuned ProtBERT-BFD
representation and a Logistic Regression classifier using the dataset explained in
Sect. 3.1, outperforms previous methods with an accuracy of 93.89% and an MCC
of 0.86 on the independent test set.
The ProtBERT-BFD representation is effective because it understands the context
of each amino acid in different protein sequences, whereas other methods rely on
static protein-encoding techniques.
Figure 3 shows a confusion matrix of TooT-BERT-T for separating transport pro-
teins from non-transport proteins. As depicted in the figure, despite the fact that the
number of errors is quite low, the model makes more mistakes when identifying non-
transporters as transporters (False positive = 6) than when predicting transporters
as non-transporters (False negative = 5). This suggests that the proposed strategy
is somewhat skewed towards predicting the positive class (transport proteins). This
issue may occur when the dataset is imbalanced, with more positive class samples
than negative class samples.
5 Conclusion
TooT-BERT-T distinguishes transport proteins from non-transport proteins using the

fine-tuned ProtBERT-BFD representation. The representations of two BERT mod-
els, ProtBERT-BFD and MembraneBERT, were compared using frozen and fine-
tuned representations. The ProtBERT-BFD fine-tuned representation outperforms the
MembraneBERT representation on the independent test set. The proposed method,
TooT-BERT-T, which utilizes fine-tuned ProtBERT-BFD and Logistic Regression,
achieves an accuracy of 93.89% and an MCC of 0.86 on the independent test set and
outperforms other methods. Given that this study was a preliminary examination of
the BERT representation’s performance in transport protein analysis, other classifiers
such as SVM and CNN can be evaluated in the future.
References
1. Alballa M, Butler G (2020) Integrative approach for detecting membrane proteins. BMC Bioin-
form 21(19):575
2. Alballa M, Butler G (2020) TooT-T: discrimination of transport proteins from non-transport
proteins. BMC Bioinform 21(3):25
3. Alley EC, Khimulya G, Biswas S, AlQuraishi M, Church GM (2019) Unified rational protein
engineering with sequence-based deep representation learning. Nat Methods 16(12):1315–
1322
4. Aplop F, Butler G (2015) On predicting transport proteins and their substrates for the recon-
struction of metabolic networks. In: 2015 IEEE conference on computational intelligence in
bioinformatics and computational biology (CIBCB), pp 1–9
5. Aplop F, Butler G (2017) TransATH: transporter prediction via annotation transfer by homol-
ogy. ARPN J Eng Appl Sci 12(2):8
6. Bepler T, Berger B (2019) Learning protein sequence embeddings using information from
structure. arXiv:1902.08661 [cs, q-bio, stat]
7. Chicco D, Jurman G (2020) The advantages of the Matthews Correlation Coefficient (MCC)
over F1 score and accuracy in binary classification evaluation. BMC Genom 21(1):6
8. Detlefsen NS, Hauberg S, Boomsma W (2022) Learning meaningful representations of protein
sequences. Nat Commun 13(1):1914
9. Devlin J, Chang MW, Lee K, Toutanova K (2019) BERT: pre-training of deep bidirectional
transformers for language understanding. arXiv:1810.04805 [cs]
10. Elnaggar A, Heinzinger M, Dallago C, Rehawi G, Wang Y, Jones L, Gibbs T, Feher T, Angerer
C, Steinegger M, Bhowmik D, Rost B (2021) ProtTrans: towards cracking the language of
lifes code through self-supervised deep learning and high performance computing. IEEE Trans
Pattern Anal Mach Intell 1
11. Ferruz N, Höcker B (2022) Towards controllable protein design with conditional transformers.
arXiv:2201.07338 [q-bio]
12. Hess AS, Hess JR (2019) Logistic regression. Transfusion 59(7):2197–2198
13. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates
R, Z̆ídek A, Potapenko A, Bridgland A, Meyer C, Kohl SAA, Ballard AJ, Cowie A, Romera-
Paredes B, Nikolov S, Jain R, Adler J, Back T, Petersen S, Reiman D, Clancy E, Zielinski M,
Steinegger M, Pacholska M, Berghammer T, Bodenstein S, Silver D, Vinyals O, Senior AW,
Kavukcuoglu K, Kohli P, Hassabis D (2021) Highly accurate protein structure prediction with
AlphaFold. Nature 596(7873):583–589
14. Leinonen R, Diez FG, Binns D, Fleischmann W, Lopez R, Apweiler R (2004) UniProt archive.
Bioinformatics 20(17):3236–3237
15. Liou YF, Vasylenko T, Yeh CL, Lin WC, Chiu SH, Charoenkwan P, Shu LS, Ho SY, Huang HL
(2015) SCMMTP: identifying and characterizing membrane transport proteins using propensity
scores of dipeptides. BMC Genom 16(12):S6
16. Mishra NK, Chang J, Zhao PX (2014) Prediction of membrane transport proteins and their
substrate specificities using primary sequence information. PLoS ONE 9(6):e100278
17. Nguyen TTD, Le NQK, Ho QT, Phan DV, Ou YY (2019) Using word embedding technique
to efficiently represent protein sequences for identifying substrate specificities of transporters.
Anal Biochem 577:73–81
18. Ofer D, Brandes N, Linial M (2021) The language of proteins: NLP, machine learning & protein
sequences. Comput Struct Biotechnol J 19:1750–1758
19. Rao R, Bhattacharya N, Thomas N, Duan Y, Chen P, Canny J, Abbeel P, Song Y (2019)
Evaluating protein transfer learning with TAPE. In: Wallach H, Larochelle H, Beygelzimer A,
Alché-Buc Fd, Fox E, Garnett R (eds) Advances in neural information processing systems,
vol 32. Curran Associates, Inc
20. Saier Jr MH, Tran CV, Barabote RD (2006) TCDB: the transporter classification database for
membrane transport protein analyses and information. Nucleic Acids Res 34(suppl_1):D181–
D186
21. Saier Jr MH (2002) Families of transporters and their classification. In: Transmembrane trans-
porters. Wiley, pp 1–17
22. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaiser L, Polosukhin I
(2017) Attention is all you need. arXiv
23. Vig J, Madani A, Varshney LR, Xiong C, Socher R, Rajani NF (2021) BERTology meets
biology: interpreting attention in protein language models. arXiv:2006.15222 [cs, q-bio]
24. Wolf T, Debut L, Sanh V, Chaumond J, Delangue C, Moi A, Cistac P, Rault T, Louf R, Funtowicz
M, Davison J, Shleifer S, von Platen P, Ma C, Jernite Y, Plu J, Xu C, Scao TL, Gugger S, Drame
M, Lhoest Q, Rush AM (2020) HuggingFace’s transformers: state-of-the-art natural language
processing. arXiv
Machine Learning and Deep Learning
Techniques for Epileptic Seizures
Prediction: A Brief Review
Marco Hernández , Ángel Canal-Alonso , Fernando de la Prieta ,

Sara Rodríguez , Javier Prieto , and Juan Manuel Corchado
Abstract The third most common neurological disorder, only behind stroke and
migraines, is Epilepsy. The main criteria for its diagnosis are the occurrence of
unprovoked seizures and the possibility of new seizures appearing. Usually, the
professional in charge of detecting these seizures is a neurologist who interprets
the patients’ electroencephalography. However, more accurate, precise, and sensitive
methods are needed. Machine learning has increased as a viable alternative, reducing
costs and ensuring rapid diagnostic time. This work reviews the state of the art in
machine learning applied to epileptic seizure detection and prediction as a prospective
study before developing a novel seizure prediction algorithm.
Keywords Seizure prediction · Machine learning · Epilepsy ·

Electroencephalogram
This work was supported by the HERMES project, funded by the European Union under the Horizon
2020 FET-proactive program, Grant Agreement n. 824164., as well as the “XAI - XAI - Sistemas
Inteligentes Auto Explicativos creados con Módulos de Mezcla de Expertos” project, ID SA082P20,
financed by Junta Castilla y León, Consejería de Educación, and FEDER funds.
M. Hernández (B) · Á. Canal-Alonso · F. de la Prieta · S. Rodríguez · J. Prieto · J. M. Corchado

BISITE Research Group, University of Salamanca, 37007 Salamanca, Spain
e-mail: marcohperez@usal.es
Á. Canal-Alonso
e-mail: acanal@usal.es
F. de la Prieta
e-mail: fer@usal.es
S. Rodríguez
e-mail: srg@usal.es
J. Prieto
e-mail: javierp@usal.es
J. M. Corchado
e-mail: corchado@usal.es

14 M. Hernández et al.
1 Introduction
Epilepsy is a neurological disorder affecting more than 39 million people in the

United States [1], being the third most common only before stroke and migraines
[31]. Due to its chronic nature, it is one of the most invalidating conditions, making
it a suitable target for new therapies and biomedical research [11, 14, 34].
The International League Against Epilepsy defines an epileptic seizure as a “tran-
sient occurrence of signs and/or symptoms due to abnormal excessive or synchronous
neuronal activity in the brain” [59].
The most widely used diagnostic method for epilepsy is electroencephalogra-
phy (EEG). EEG records the electrical activity of the brain by the voltage changes
provoked by the ion currents of the brain neurons [44].
These voltage changes are recorded with electrodes placed either along the scalp
(sEEG) or over the cortex, inside the cranium (iEEG). Those electrodes do not have
enough spatial resolution to detect each neuron’s action potential, but the simul-
taneous activity of millions of neurons creates a voltage wave that stimulates the
electrodes.
The events observed in the EEG can be divided into three categories [16]: Ictal
events, which are the events occurring during the seizure, preictal events, the ones
that preceded the seizure, and interictal events, every other event that is not part of
the ictal or preictal phase. The duration of the preictal phase varies between studies,
and can range from a few minutes to hours.
In animal and in vitro models, microelectrodes are used instead of EEG to per-
form electrophysiology recordings, as the latter carries a considerable amount of
drawbacks in such cases (Fig. 1).
2 Signal Processing and Feature Extraction
EEG signals usually come with several artifacts that may obscure the signal’s
epilepsy-related information. These artifacts depend on the type of model of study
or the specifics of the EEG recording.
Fig. 1 Usual pipeline in epileptic seizure classification or prediction

ML and DL Techniques for Epileptic Seizures Prediction ... 15
The most typical processing for most EEG signals includes the removal of back-
ground noise, done by filtering the 50–60 Hz powerline. A vast number of features
can then be extracted from the processed signal.
2.1 Feature Extraction
Statistical Features. Distribution and amplitude changes in EEG signals can be

tracked using statistical parameters like kurtosis, mean, skewness, and variance [12,
38, 53]. Phase correlation [39, 46, 49, 57] can be used to analyze the patterns in ictal
and preictal events. Common Spatial Pattern [56] extracts features by decomposing
EEG signals.
Nonlinear Features. Correlation Dimension allows to measure the complexity of
each event [2, 27] and the Largest Lyapunov Exponent calculates the chaos in EEG
signals [60, 61]. Applying Fractality Dimension [17, 18, 41] enables the comparison
of the rhythms between the EEG events and exposes the self-similarity in the data.
The Repeatability of the events is also measurable using Lempel-Zic Complexity [2,
10, 71]and Approximate Entropy [58, 70]. Entropy can also be used as a measure
of randomness with Spectral Entropy and level of disorganization using Wavelet
Entropy.
Activity (variance of the signal of a time function), mobility (proportion of the
standard deviation), and complexity (change in frequency) parameters can be used
as descriptors in the Hjorth parameters analysis [19, 32]. To unveil the unifor-
mity of the different frequency bands in EEG data Wavelet Energy can be applied
[6, 24, 26].
Frequency Domain Features. Fourier transforms like Short Time Fourier Transform
[63] and Fractional Fourier Transform [48] are used to obtain phase and magnitude
components, while Spectral Power Analysis [7, 21] allows studying the different
frequency bands in EEG.
Time-Frequency Domain Features. One of the most widely used time-frequency
domain feature in EEG analysis is the wavelet transform, either Discrete Wavelet
Transform [68], or Continuous Wavelet Transform [40]. Each wavelet transform
offers a different decomposition; the CWT generates a scalogram from the dilation
and translation, while the DWT filters the signal and breaks down the signal in
different levels.
Some time-frequency domain features can be combined with nonlinear features to
monitor hidden data properties. Higher-Order Spectra [4, 45] and Variational Model
Decomposition [20, 35] are the most widely used features in these combinations.
2.2 Feature Selection
For most Machine Learning techniques, selecting an optimum number of features is

essential for the algorithms to reach their full potential. Having features that carry
similar information about the target variable or no information about this variable
whatsoever can make the model too complex and impoverish its performance.
A roster of techniques is used to select the most suitable features for each study.
Statistical approaches like Principal Components Analysis [23, 50, 66] or Partial
Least Squares [30] are the easiest way of archiving a conclusion. More complex
techniques such as Minimum Redundancy Maximum Relevance [8, 51] or Gaussian
Mixture Models [22, 52, 69] are used in more sensitive situations like human seizure
prediction.
However, Deep Learning algorithms can use part of their architecture to automat-
ically extract the most relevant information out of the initial input variables [65, 67].
In these cases, both feature extraction and selection can be skipped and still achieve
competitive results.
3 Seizure Detection and Classification
Detection of seizures by EEG has traditionally been done manually by clinical pro-
fessionals, evaluating the frequency, wavelength, voltage, amplitude, and waveforms.
These features are suitable for being analyzed using automated learning algorithms
[28]. Since the first computer analysis of EEG records in 2002 using wavelet trans-
form [5], automated detection of ictal events has become an essential matter in
epilepsy research [9, 25].
Many studies have been carried out in the last years to improve the performance of
automated seizure detection. A wide range of Machine Learning and Deep Learning
classifiers have been employed [15, 37] but, while detecting seizures may remain
valuable for research purposes, patients and clinical professionals need tools that
allow them to avoid the seizures instead of doing a post hoc analysis; here is where
seizure prediction comes necessary.
4 Seizure Prediction
Having enough anticipation before a seizure is a crucial milestone for a clinical

approach. The usual pipeline to seizure prediction is similar to seizure detection.
However, instead of classifying interictal and ictal events, it focuses on separating
interictal from preictal, leaving the actual seizures out of the analysis.
Establishing a correct stimulation protocol implies having enough anticipation

before the seizure. Nevertheless, early prediction usually reduces sensitivity and
specificity, ballasting the overall performance [29, 42].
4.1 Animal Models
The first experiments achieved times near 2.24 min and demonstrated the effective-
ness of wavelet functions as predictors [47]. Best time results in animal models were
obtained by [64] using canine EEG data and multiple machine learning algorithms,
being able to detect the seizures 1 min ahead. That work established a proof of con-
cept, so no diagnostic performance analyses were carried out. Nevertheless, some
works have developed systems with performances over 90% of sensitivity but with a
loose prediction time. Rajdev’s team [55] developed a seizure prediction system for
rat EEG recordings based on an adaptive wiener filter; his approach hits a 92% of
sensitivity, being also the most sensitive of the works done on animal EEG recordings.
4.2 Human Subjects
The work of Iasemidis [33] established a ceiling of capacity in the time of prediction
with 91 min and a precision of 91.3% (and sensitivity of 81.82%). Other authors
aimed to maintain a prediction time horizon and keep the algorithm between those
parameters. In such works [2, 70] 30 and 50-min horizons were fixed, and sensitivity
between 79.9 and 90.2% were achieved.
Tsiouris [62] reached a prediction with 15 to 120 min ahead and a 99% of sensitiv-
ity, making use of Long Short-term Memory networks (LSTM), the first application
of deep learning in the field.
Other deep learning approaches have reached similar results while automating
feature extraction. Wei [65] uses an image of the EEG as input to an architecture
based on Convolutional Neural Networks (CNN) for feature extraction and LSTM
for sequence learning. This network achieves an average accuracy of 93.4% at an
average warning time of 21 min. Transformers have been used in a similar manner
[67] reaching prediction sensitivity and a False Positives Rate of 96.01% and 0.047/h,
respectively with an average warning time between 3 and 30 min.
5 Conclusion
Although many advances have been made since the first works in automated seizure
prediction, some gaps remain to be cleared.
To deploy accurate diagnostic and treatment tools, a correct balance between

the time of prediction and the sensitivity and specificity must be reached. This bal-
ance can only be achieved with deep learning techniques such as CNN, LSTM, and
Transformers. In this regard, some recent developments have been made, reaching
promising results [3, 13, 36, 43, 54, 65, 67].
A combination of multiple techniques and features offers the best performance
and results, but the computational requirements scale with each model implemented.
Solving this issue is also a significant challenge in automated seizure prediction.
The solution to the automated seizure prediction problem will surely enhance the
life quality of epilepsy patients and ameliorate the impact on the health services.
References
1. Global, regional, and national incidence, prevalence, and years lived with disability for 310
diseases and injuries, 1990-2015: a systematic analysis for the global burden of disease study.
Technical report, GBD 2015 Disease and Injury Incidence and Prevalence Collaborators (2015)
2. Aarabi A, He B (2012) A rule-based seizure prediction method for focal neocortical epilepsy.
Clin Neurophysiol 123:1111–1122
3. Acharya UR, Oh S, Hagiwara Y, Adeli H (2017) Deep convolutional neural network for the
automated detection of seizure using EEG signals. Comput Biol Med 100:270–278
4. Acharya U, Yanti R, Zheng J, Mookiah M, Tan J, Martis R et al (2013) Automated diagnosis
of epilepsy using CWT, HOS and texture parameters. Int J Neural Syst 23
5. Adeli H, Zhou Z, Dadmehr N (2003) Analysis of EEG records in an epileptic patient using
wavelet transform. J Neurosci Methods 123:69–87
6. Alves N, Rodrigues R, Rocha M (2022) BioTMPy: a deep learning-based tool to classify
biomedical literature. In: Lecture notes in networks and systems. LNNS, vol 325, pp 115–125
7. Bandarabadi M, Rasekhi J, Teixeira C, Karami M (2015) On the proper selection of preictal
period for seizure prediction. Epilepsy Behav 45:158–166
8. Bandarabadi M, Teixeira C, Rasekhi J, Dourado A (2015) Epileptic seizure prediction using
relative spectral power features. Clin Neurophysiol 126:237–248
9. Calvaresi D, Albanese G, Calbimonte J, Schumacher M (2020) Seamless: simulation and anal-
ysis for multi-agent system in time-constrained environments. In: Lecture notes in computer
science (including subseries lecture notes in artificial intelligence and lecture notes in bioin-
formatics). LNAI, vol 12092, pp 392–397
10. Casado-Vara R, González-Briones A, Prieto J, Corchado J (2019) Smart contract for monitoring
and control of logistics activities: pharmaceutical utilities case study. Adv Intell Syst Comput
771:509–517
11. Casado-Vara R, Novais P, Gil A, Prieto J, Corchado J (2019) Distributed continuous-time fault
estimation control for multiple devices in IoT networks. IEEE Access 7:11972–11984
12. Casado-Vara R, Prieto-Castrillo F, Corchado J (2018) A game theory approach for cooperative
control to improve data quality and false data detection in WSN. Int J Robust Nonlinear Control
28(16):5087–5102
13. Chen H, Shen J, Wang L, Jin Y (2021) Towards a more effective bidirectional LSTM-based
learning model for human-bacterium protein-protein interactions. In: Advances in intelligent
systems and computing. AISC, vol 1240, pp 91–101
14. Costa Â, Novais P, Corchado J, Neves J (2012) Increased performance and better patient
attendance in an hospital with the use of smart agendas. Logic J IGPL 20(4):689–698
15. Cristani M, Tomazzoli C, Olivieri F, Pasetto L (2020) An ontology of changes in normative
systems from an agentive viewpoint. In: Communications in computer and information science.
CCIS, vol 1233, pp 131–142
16. Cámpora NE, Mininni CJ, Kochen S, Lew SE (2019) Seizure localization using pre ictal phase-
amplitude coupling in intracranial electroencephalography. Sci Rep 9:20022
17. D’Alessandro M, Esteller R, Vachtsevanos G, Hinson A, Echauz J, Litt B (2003) Epileptic
seizure prediction using hybrid feature selection overmultiple intracranial EEG electrode con-
tacts: a report of four patients. IEEE Trans Inf Theory 50:603–615
18. De Meo P, Falcone R, Sapienza A (2020) Fast and efficient partner selection in large agents’
communities: when categories overcome direct experience. In: Lecture notes in computer
science (including subseries lecture notes in artificial intelligence and lecture notes in bioin-
formatics). LNAI, vol 12092, pp 106–117
19. Direito B, Teixeira CA, Sales F, Castelo-Branco M, Dourado A (2017) A realistic seizure
prediction study based on multiclass SVM. Int J Neural Syst 27:1750006–1750021
20. Dragomiretskiy K, Zosso D (2014) Variational mode decomposition. IEEE Trans Signal Pro-
cess 62:531–544
21. Dressler O, Schneider G, Stockmanns G, Kochs E (2004) Awareness and the EEG power
spectrum: analysis of frequencies. Br J Anaesth 93
22. D’Auria M, Scott E, Lather R, Hilty J, Luke S (2020) Assisted parameter and behavior calibra-
tion in agent-based models with distributed optimization. In: Lecture notes in computer science
(including subseries lecture notes in artificial intelligence and lecture notes in bioinformatics).
LNAI, vol 12092, pp 93–105
23. Fatima N (2020) Enhancing performance of a deep neural network: a comparative analysis of
optimization algorithms. ADCAIJ Adv Distrib Comput Artif Intell J 9(2):79–90
24. Gadhoumi K, Lina J, Gotman J (2013) Seizure prediction in patients with mesial temporal lobe
epilepsy using EEG measures of state similarity. Clin Neurophysiol 124:1745–1754
25. García-Retuerta D, Canal-Alonso A, Casado-Vara R, Rey A, Panuccio G, Corchado J (2021)
Bidirectional-pass algorithm for interictal event detection. In: Advances in intelligent systems
and computing. AISC, vol 1240, pp 197–204
26. Garg G, Singh V, Gupta J, Mittal A (2011) Relative wavelet energy as a new feature extractor
for sleep classification using EEG signals. Int J Biomed Signal Process 2:75–80
27. Grassberger P, Procaccia I (1983) Characterization of strange attractors. Phys Rev Lett 50:346–
349
28. Gupta S, Meena J, Gupta O (2020) Neural network based epileptic EEG detection and classi-
fication. ADCAIJ Adv Distrib Comput Artif Intell J 9(2):23–32
29. Gupta S, Ranga V, Agrawal P (2022) EpilNet: a novel approach to IoT based epileptic seizure
prediction and diagnosis system using artificial intelligence. ADCAIJ Adv Distrib Comput
Artif Intell J 10(4):435–452
30. Haenlein A (2004) A beginner’s guide to partial least squares analysis. Underst Stat 283–297
31. Hirtz D, Thurman DJ, Gwinn-Hardy K, Mohamed M, Chaudhuri AR, Zalutsky R (2007) How
common are the “common” neurologic disorders? Neurology 68:326–337
32. Hjorth B, Elema-Schonander A (1970) EEG analysis based on time domain properties. Elec-
troencephalogr Clin Neurophysiol 29:306–310
33. Iasemidis L, Shiau D, Pardalos P, Chaovalitwongse W, Narayanon K, Prasad A (2005) Long-
termprospective on-line real-time seizure prediction. Clin Neurophysiol 116:532–544
34. Khan A, Zubair S, Khan S (2021) Comprehensive performance analysis of neurodegenerative
disease incidence in the females of 60–96 year age group. ADCAIJ Adv Distrib Comput Artif
Intell J 10(2):183–196
35. Kumar M, Rao Y (2018) Epileptic seizures classification in EEG signal based on semantic
features and variational mode decomposition. Clust Comput 1–11
36. Lane N, Kahanda I (2021) DeepACPpred: a novel hybrid CNN-RNN architecture for predicting
anti-cancer peptides. In: Advances in intelligent systems and computing. AISC, vol 1240, pp
60–69
37. Lee K, Jeong H, Kim S, Yang D, Kang HC, Choi E (2022) Real-time seizure detection
using EEG: a comprehensive comparison of recent approaches under a realistic setting.
arXiv:2201.08780 [cs]
38. Li T, Fan H, García J, Corchado J (2018) Second-order statistics analysis and comparison
between arithmetic and geometric average fusion: application to multi-sensor target tracking.
Inf Fusion 51:233–243
39. Li T, Su J, Liu W, Corchado J (2017) Approximate gaussian conjugacy: parametric recursive
filtering under nonlinearity, multimodality, uncertainty, and constraint, and beyond. Front Inf
Technol Electron Eng 18(12):1913–1939
40. Mallat S, Hwang W (1992) Singularity detection and processing with wavelets. IEEE Trans
Inf Theory 38:617–643
41. Mandelbrot B (1983) Geometry of nature. Freeman
42. Mena Mamani N (2020) Machine learning techniques and polygenic risk score application to
prediction genetic diseases. ADCAIJ Adv Distrib Comput Artif Intell J 9(1):5-14
43. Muhamada AW, Mohammed AA (2022) Review on recent computer vision methods for human
action recognition. ADCAIJ Adv Distrib Comput Artif Intell J 10(4):361–379
44. Niedermeyer E, da Silva F (2004) Electroencephalography: basic principles, clinical applica-
tions, and related fields. Williams & Wilkins
45. Nikias C, Petropulu A (1993) Higher order spectra analysis: a nonlinear signal processing
framework. PTR Prentice Hall
46. Nugroho S, Weinmann A, Schindelhauer C, Christ A (2020) Averaging emulated time-series
data using approximate histograms in peer to peer networks. In: Communications in computer
and information science. CCIS, vol 1233, pp 339–346
47. Ouyang G, Li X, Guan X (2007) Application of wavelet-based similarity analysis to epileptic
seizures prediction. Comput Biol Med 37:430–437
48. Ozaktas H, Zalevsky Z, Kutay M (2001) The fractional Fourier transform. Wiley
49. Parvez M, Paul M (2016) Epileptic seizure prediction by exploiting spatiotemporal relationship
of EEG signals using phase correlation. IEEE Trans Neural Syst Rehabil Eng 24:158–168
50. Pearson K (1901) On lines and planes of closest fit to systems of points in space. Philos Mag
2:559–572
51. Peng H, Long F, Ding C (2005) Feature selection based on mutual information: criteria of
max-dependency, max-relevance, and min-redundancy. IEEE Trans Pattern Anal Mach Intell
27:1226–1238
52. Press M (ed) (1999) The infinite Gaussian mixture model
53. Press W, Flannery B, Teukolsky S, Vetterling W (1992) Numerical recipes in C: the art of
scientific computing. Cambridge University Press
54. Pérez-López R, Blanco G, Fdez-Riverola F, Lourenço A (2021) The activity of bioinformatics
developers and users in stack overflow. In: Advances in intelligent systems and computing.
AISC, vol 1240, pp 23–31
55. Rajev P, WArd M, Rickus J, Worth R, Irazoqui P (2010) Real-time seizure prediction from
local field potentials using an adaptive wiener algorithm. Comput Biol Med 40:97–108
56. Ramoser H, Mller-Gerking J, Pfurtscheller G (2000) Optimal spatial filtering of single trial
EEG during imagined hand movement. IEEE Trans Rehabil Eng 8:441–446
57. Reddy B, Chatteriji B (1996) An FFT-based technique for translation, rotation, and scale
invariant image registration. IEEE Trans Image Process 5:1266–1271
58. Richman J, Moorman J (2000) Physiological time-series analysis using approximate entropy
and sample entropy. Am J Physiol Heart Circ Physiol 278:2039–2049
59. Robert S, Fisher AC, Arzimanoglou A, Bogacz A, Cross JH, Elger Jr CE, Forsgren L, French
JA, Glynn M, Hesdorffer DC, Lee B, Mathern GW, Moshé SL, Perucca E, Scheffer IE, Tomson
T, Watanabe M, Wiebe S (2014) ILAE official report: a practical clinical definition of epilepsy.
Epilepsia 55:475–482
60. Rosenstein M, Colins J, de Luca C (1993) A practical method for calculating largest Lyapunov
exponents from small data sets. Phys D 65:117–134
61. Shafique A, Sayeed M, Tsakalis K (2018) Nonlinear dynamical systems with chaos and big
data: a case study of epileptic seizure prediction and control. Guide to big data applications.
Springer (2018)
62. Tiouris K, Pezoulas V, Zervakis M, Konitsiotis S, Koutsouris D, Fotiadis D (2018) A long

short-term memory deep learning network for the prediction of epileptic seizures using EEG
signals. Comput Biol Med 99:24–37
63. Truong N, Nguyen A, Kuhlmann L, Bonyadi M, Yang J, Ippolito S et al (2018) Convolutional
neural networks for seizure prediction using intracranial and scalp electroencephalogram. Neu-
ral Netw 105:104–111
64. Varatharajah Y, Iyer R, Berry B, Worrell G, Brinkmann B (2017) Seizure forecasting and the
preictal state in canine epilepsy. Neural Syst 27
65. Wei X, Zhou L, Zhang Z, Chen Z, Zhou Y (2019) Early prediction of epileptic seizures using
a long-term recurrent convolutional network. J Neurosci Methods 327:108395
66. Williamson J, Bliss D, Browne D, Narayanan J (2012) Seizure prediction using EEG spa-
tiotemporal correlation structure. Epilepsy Behav 25:230–238
67. Yan J, Li J, Xu H, Yu Y, Xu T (2022) Seizure prediction based on transformer using scalp
electroencephalogram. Appl Sci 12:4158
68. Yves M (1992) Wavelets and operators. Cambridge University Press
69. Zandi A, Tafreshi R, Javidan M, Dumont G (2013) Predicting epileptic seizures in scalp EEG
based on a variational Bayesian Gaussian mixture model of zero-crossing intervals. IEEE Trans
Biomed Eng 60:1401–1413
70. Zhang Y, Zhou W, Yuan Q, Wu Q (2014) A low computation cost method for seizure prediction.
Epilepsy Res 108:1357–1366
71. Ziv J, Lempel A (1977) A universal algorithm for sequential data compression. IEEE Trans
Inf Theory 23:337–343
The Covid-19 Decision Support System
(C19DSS) – A Mobile App
Pierpaolo Vittorini , Nicolò Casano , Gaia Sinatti ,

Silvano Junior Santini , and Clara Balsano
Abstract The COVID-19 pandemic remains a concrete challenge, especially in

communities and rural areas where health resources are scarce. We recently devel-
oped several classifiers, useful to predict safe discharge, disease severity, and mor-
tality risk from COVID-19, fed by routine analyses collected in the Emergency
Department. In this paper, we discuss a system, made up of an app and a server, that
enables doctors to use these models during the management of COVID-19 patients.
The app has been developed involving the doctors since the early phases of the app
design, then revised in the light of two usability cycles. We report its main features
and its ease of use. So far, it has been used during the fourth wave, producing accurate
results with patients that did not complete the vaccination protocol (i.e., up to the
second dose).
Keywords COVID-19 · App · Machine learning · User-centered design
1 Introduction
Health informatics can be defined as the application of computer science, engineering

and telecommunication to healthcare [2]. It regards the use of methods, applications
and devices in all aspects concerned with both individuals and public health [8, 11,
23, 25].
The pandemic caused by severe respiratory acute syndrome coronavirus 2 (SARS-
CoV-2) was declared a global emergency by the World Health Organization (WHO)
[24] on the 11th of March 2020. Although recent progress in the possible treatments
has changed the face of the pandemic, some concerns still remain about threats
related to SARS-CoV-2 [21]. Countries have not adopted a common global response
to COVID-19 and vaccination inequities are manifest [17]. In this scenario, the
pandemic risk remains a concrete challenge, especially in communities and rural
P. Vittorini (B) · N. Casano · G. Sinatti · S. J. Santini · C. Balsano

University of L’Aquila, L’Aquila 67100, Italy
e-mail: pierpaolo.vittorini@univaq.it
URL: https://vittorini.univaq.it/

24 P. Vittorini et al.
areas where health resources are scarce. To face these risks, we recently developed
several classifiers, useful to predict safe discharge, disease severity, and mortality risk
from COVID-19, fed by routine analyses collected in the Emergency Department
(ED) [7].
In this paper, we focus on a system, called COVID-19 Decision Support System
(C19DSS), whose aim is to enable doctors from EDs to take advantage of the afore-
mentioned models during the management of COVID-19 patients. The C19DSS sys-
tem has been developed following the User-Centered Design (UCD) methodology,
i.e., involving the doctors since the early phases of the system design, and revising
the development in the light of usability cycles [16]. So far, the system has been used
during the fourth wave. It is currently producing accurate results with patients that
did not complete the vaccination protocol (i.e., up to the second dose). Therefore,
we consider the system suitable in these cases, as well as in all countries with low
vaccination rates.
2 Background
To allow the paper to be self-contained, we briefly report the results concerning the
models we developed, which we submitted in [7].
From a dataset containing the routine analyses of 779 patients collected in the
ED, we devised several models for both the complete cases and through missing
data imputation [6]. The following different models were tried from the available
dataset: decision tree (DT) – as baseline [5], random forest (RF) [4] and gradient
boosting machines (GBM) [9]. The models were developed to predict safe discharge
(discharge/admit), disease severity (mild/severe), and mortality risk (no risk/risk).
For all models and outcomes, we split the dataset into train and test (with 75% of
data going for training, 25% for testing), used 10-fold cross-validation, tuned each
classifier according to its specific hyper-parameters, calculated the confusion matrix
and the ROC curve [10]. The results are summarized in Table 1. The table lists all
details of each model, for each outcome, for the subset of the complete cases and
for the complete dataset (with missing data imputation). The best AUC is reported
in bold, together with the corresponding model.
Besides the limited size of the dataset and the constraint of using only routine
clinical and laboratory data to devise the models, the performances of our models
are in line with the best prediction models available in the scientific literature, that
make use of similar data than our [3, 12, 26]. In particular: (i) concerning hospital
admission, Jimenez-Solem et al. [12] developed a RF model that reached an AUC
equal to 0.82, vs 0.89 and 0.94 of our models; (ii) for severity prediction, Yao et
al. [26] devised a Support Vector Machines (SVM) model that reached an accuracy
equal to 0.82, vs the 0.76 and 0.89 of accuracy of our models; (iii) for mortality
prediction, Booth et al. [3] developed a SVM model that reached an AUC of 0.93,
vs 0.84 and 0.87 of our models.
C19DSS 25
Table 1 Main statistics for all outcomes and classifiers. Acc = Accuracy, Sens = Sensitivity, Spec
= Specificity, AUC = Area Under the Curve
Complete cases Missing data imp.
Acc Sens Spec AUC Acc Sens Spec AUC
Safe discharge
DT 0.870 0.941 0.546 0.937 0.824 0.842 0.770 0.858 DT
RF 0.886 0.960 0.546 0.938 0.829 0.869 0.780 0.894 RF
GBM 0.886 0.951 0.591 0.943 0.824 0.876 0.666 0.882 GBM
Disease severity
DT 0.805 0.560 0.867 0.792 0.742 0.732 0.752 0.766 DT
RF 0.886 0.680 0.939 0.886 0.757 0.783 0.732 0.832 RF
GBM 0.846 0.600 0.908 0.893 0.762 0.804 0.721 0.827 GBM
Mortality
DT 0.829 0.970 0.250 0.758 0.840 0.944 0.290 0.689 DT
RF 0.878 0.960 0.542 0.866 0.876 0.969 0.387 0.842 RF
GBM 0.854 0.939 0.500 0.857 0.860 0.944 0.419 0.844 GBM
With respect to similar tools available in the scientific literature, Liu et al. [15]
propose a system to assist doctors in collecting data, assessing risk, triaging, manag-
ing, and following up on patients during the COVID-19 outbreak. The system uses
logistic regression to predict risk, obtaining an AUC of 0.71. Furthermore, McRae
et al. [18] developed an app that leverages models that use non-laboratory data to
help determine whether hospitalization is necessary (AUC = 0.79) and that predicts
the probability of mortality using bio-marker measurements (AUC = 0.95).
Our work continues in the same direction: it adopted state-of-the-art models based
on routine data collected by the involved EDs, and developed a system supporting
doctors in defining the need for hospitalization, disease severity and mortality risk,
for patients accessing the ED.
3 C19DSS
In this section, we first describe the architecture of the overall system, and in particular
of the C19CDSS app. Then, we present the usability results and the preliminary use
within our Institution.
Fig. 1 System architecture
3.1 Architecture
The COVID-19 Decision Support System (C19DSS) is the system we developed

to enable physicians to effectively use our models. The system is made up of a
smartphone app used by clinicians, and a server that provides the “intelligence” to
the app (Fig. 1).
The app is made up of four activities (see Fig. 2). The first activity is the dashboard,
where a summary of the database and of the server connection status is reported. The
second screen contains the patient list, how to filter patients according to different
parameters, and the button to add a new patient. The third screen shows how to
enter the laboratory/clinical data of a new patient, and the button to request the
classification to the server. The fourth screen depicts how to edit the patient data,
request the classification to the server, or delete the patient from the database.
In details, the first screen is the app dashboard (Fig. 2-a). The first card contains
a summary of the data stored in the database. The second card shows the connection
status with the server. By tapping on the “Go”, the user accesses the list of available
patients (Fig. 2-b). If needed, the user can show a filtering panel through a menu item
“Filters”. The filtering panel permits to include/exclude patients: (i) that have been
classified and/or finalized (i.e., whose hospitalization has ended), or (ii) that have a
name/surname containing a given string. The central part of the screen contains the
list of patients (with information about name, surname, ID and status): by tapping on
a patient, the user can edit the related data; by tapping on the floating action button, the
user can add a new patient to the database. Figure 2-c shows the interface to introduce
all data regarding a patient. On the lower part of the interface, two floating action
buttons are available. The rightmost button saves the data, the leftmost requests the
classification. Finally, Fig. 2-d shows how the user can edit the patient’s data. Three
floating action buttons are available. From right to left, allow a user to update the
data, request the classification, and delete the patient. Worth noting the two panels
at the bottom of the interface, i.e., “Automated classification” and “Outcome”, that
contain the results of the automated classification and the hospitalization outcome,
respectively.
When the classification process is activated, the app opens an encrypted con-
nection to the server, sends the laboratory data and the ID of the patient (so, no
C19DSS 27
(a) Dashboard (b) Patient list (c) New data (d) Edit data
Fig. 2 C19DSS activities
personal data is ever communicated over the network) to the classification endpoint
(the server follows the RESTful API paradigm) [20]. Then, the server uses R [19] to
apply the correct model, depending on the request, on the received data. Hence, the
server stores the received data for further analyses (i.e., to evaluate the quality of the
predictions and potentially update the models), and finally returns the classification
results to the app.
3.2 Usability Evaluation
To develop the app, we followed the UCD methodology, i.e., we involved the physi-
cians from the very beginning phases of the design, and then we adapted and improved
the design/implementation according to consecutive cycles of usability tests.
In the first phase, we discussed and defined with three physicians the navigational
structure and the app user interface through mockups. After the system implementa-
tion, the first usability testing took place. The following three tasks were evaluated
with seven physicians: (i) data entry, (ii) classification and (iii) data editing. We
collected quantitative and qualitative measures based on the Single Ease Question
(SEQ) and through unstructured interviews [22]. At the first iteration, we measured
an average SEQ of 3.86/5, 3.71/5 and 4.00/5 for each task, and we collected a few
issues and suggestions on how to improve the app. Among them, we added the auto-
mated calculation of the P/F, NLR and PLR values1 , we implemented a more clear
1 P/F (PaO2 /FIO2 ) = Oxygenation Index, NLR = Neutrophil-to-Lymphocyte Ratio, PLR = Platelet-
to-Lymphocyte Ratio.
visualization of the classification, and we fixed a bug that blocked the classification.
At the second iteration, the average SEQs increased to 4.71/5, 4.43/5 and 4.71/5.
In summary, we increased the overall average ease of completing all tasks from
3.86/5 to 4.62/5, from the first to the second implementation.
3.3 Preliminary Use
So far, the system is currently used in our Institution, as complementary support to

physicians during the management of patients affected by COVID-19. The physicians
that used the system reported that the application was easy ed intuitive to use; the
process of data entry and classification did not hamper the normal ED work routine.
Conversely, it helped them to organize the workflow of COVID-19 patients.
Furthermore, with the current small cohort of patients managed through the
C19DSS system, the mortality risk prediction model showed an accuracy of 0.92,
whereas the model about safe discharge returned an accuracy of 0.57 (0.70 for the
unvaccinated cohort). However, for safe discharge, the mistakes were conservative,
i.e., the system never suggested discharging a patient that needed to be hospitalized,
and took place mostly on vaccinated patients.
4 Discussion
The work presented in this paper starts from previous research finalized to devise
state-of-the-art ML models, fed by routine clinical and laboratory analyses, to be
used by physicians to manage safe discharge, severe disease (on the seventh day
after medical presentation) and mortality during hospitalization.
Nevertheless, the models were devised from a cohort of unvaccinated patients,
hence a cohort not previously immunized against SARS-CoV-2, and therefore the
applicability of the models should be considered for unvaccinated patients.
At the time of writing, available data suggest long-term vaccine effectiveness in
fully vaccinated healthy adults, but there are some uncertainties regarding vaccine
waning in not fully vaccinated and in immunocompromised patients. Some evidence
suggests that the risk of severe disease is higher in immunocompromised patients and
in elderly ones [1, 13, 14]. On these bases, the app could be useful also in vulnerable
patients where the immunizations seem to be less effective after a prolonged time.
In order to optimize the app performance also in fully vaccinated patients, during
the data entry, for any new patient, we also save the vaccination status. So far, this
information is not used by our models. However, when enough data will be collected,
we could devise new models that will also consider the vaccination status. Moreover,
given the client/server architecture and given that the predictions are provided by the
server, the new models could be used by physicians without any change in the app,
but only with a server upgrade, without affecting the user experience.
C19DSS 29
With specific regard to the C19DSS system, the adoption of the UCD methodology
to design and develop the app, enabled us to gradually improve the user experience
and collect useful suggestions on how to improve the overall system. Finally, the
physicians that used the system reported that the application was easy ed intuitive to
use; the process of data entry and classification did not hamper the normal ED work
routine; conversely, it helped them to organize the workflow of COVID-19 patients.
5 Conclusions
Presumably, in the next future, the SARS-CoV-2 pandemic will no longer be a global
emergency, but in absence of an efficient global vaccination campaign, SARS-CoV-2
outbreaks could still be a threat to the communities where healthcare resources are
limited and the immunization rate has not reached a protective stage.
Our study highlighted how AI-powered tools could be a valid support for emer-
gency care. We do not suppose that mobile apps could replace the physician’s bedside
decision process, but we conceive that the interaction between emergency physicians
and AI tools could improve healthcare assistance and have a significant impact on
SARS-CoV-2 management.
References
1. Andrews N et al (2022) Duration of protection against mild and severe disease by COVID-19
vaccines. N Engl J Med 386(4):340–350. https://doi.org/10.1056/NEJMOA2115481
2. Bath PA (2008) Health informatics: current issues and challenges. J Inf Sci 34(4):501–518.
https://doi.org/10.1177/0165551508092267
3. Booth AL, Abels E, McCaffrey P (2021) Development of a prognostic model for mortality in
COVID-19 infection using machine learning. Mod Pathol 34(3):522–531. an official journal of
the United States and Canadian Academy of Pathology, Inc. https://doi.org/10.1038/S41379-
020-00700-X
4. Breiman L (2001) Random forests. Mach Learn 45(1):5–32.
https://doi.org/10.1023/A:1010933404324
5. Breiman L, Friedman JH, Olshen RA, Stone CJ (2017) Classification and Regression Trees.
CRC Press, Boca Raton. https://doi.org/10.1201/9781315139470
6. Van Buuren S, Groothuis-Oudshoorn K (2011) MICE: multivariate imputation by chained
equations in R. J Stat Softw 45(3):1–67. https://doi.org/10.18637/JSS.V045.I03
7. Casano N et al (2022) Application of machine learning approach in Emergency Department
to support clinical decision making for SARS-CoV-2 infected patients. Submitted manuscript,
under review
8. Chamoso P, De La Prieta F, Eibenstein A, Santos-Santos D, Tizio A, Vittorini P (2017) A device
supporting the self management of tinnitus. In: Rojas I, Ortuño F (eds) IWBBIO 2017, vol
10209. LNCS. Springer, Cham, pp 399–410. https://doi.org/10.1007/978-3-319-56154-7_36
9. Chen T, Guestrin C (2016) XGBoost: a scalable tree boosting system. In: Proceedings of the
22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp
785–794. ACM, New York, NY, USA. https://doi.org/10.1145/2939672
Another random document with
no related content on Scribd:
THE SICK CHAMBER
The New Monthly Magazine.]

[August, 1830.
What a difference between this subject and my last—a ‘Free

Admission!’ Yet from the crowded theatre to the sick chamber, from
the noise, the glare, the keen delight, to the loneliness, the darkness,
the dulness, and the pain, there is but one step. A breath of air, an
overhanging cloud effects it; and though the transition is made in an
instant, it seems as if it would last for ever. A sudden illness not only
puts a stop to the career of our triumphs and agreeable sensations,
but blots out and cancels all recollection of and desire for them. We
lose the relish of enjoyment; we are effectually cured of our romance.
Our bodies are confined to our beds; nor can our thoughts wantonly
detach themselves and take the road to pleasure, but turn back with
doubt and loathing at the faint, evanescent phantom which has
usurped its place. If the folding-doors of the imagination were
thrown open or left a-jar, so that from the disordered couch where
we lay, we could still hail the vista of the past or future, and see the
gay and gorgeous visions floating at a distance, however denied to
our embrace, the contrast, though mortifying, might have something
soothing in it, the mock-splendour might be the greater for the actual
gloom: but the misery is that we cannot conceive any thing beyond or
better than the present evil; we are shut up and spell-bound in that,
the curtains of the mind are drawn close, we cannot escape from ‘the
body of this death,’ our souls are conquered, dismayed, ‘cooped and
cabined in,’ and thrown with the lumber of our corporeal frames in
one corner of a neglected and solitary room. We hate ourselves and
everything else; nor does one ray of comfort ‘peep through the
blanket of the dark’ to give us hope. How should we entertain the
image of grace and beauty, when our bodies writhe with pain? To
what purpose invoke the echo of some rich strain of music, when we
ourselves can scarcely breathe? The very attempt is an impossibility.
We give up the vain task of linking delight to agony, of urging torpor
into ecstasy, which makes the very heart sick. We feel the present
pain, and an impatient longing to get rid of it. This were indeed ‘a
consummation devoutly to be wished’: on this we are intent, in
earnest, inexorable: all else is impertinence and folly; and could we
but obtain ease (that Goddess of the infirm and suffering) at any
price, we think we could forswear all other joy and all other sorrows.
Hoc erat in votis. All other things but our disorder and its cure seem
less than nothing and vanity. It assumes a palpable form; it becomes
a demon, a spectre, an incubus hovering over and oppressing us: we
grapple with it: it strikes its fangs into us, spreads its arms round us,
infects us with its breath, glares upon us with its hideous aspect; we
feel it take possession of every fibre and of every faculty; and we are
at length so absorbed and fascinated by it, that we cannot divert our
reflections from it for an instant, for all other things but pain (and
that which we suffer most acutely,) appear to have lost their pith and
power to interest. They are turned to dust and stubble. This is the
reason of the fine resolutions we sometimes form in such cases, and
of the vast superiority of a sick bed to the pomps and thrones of the
world. We easily renounce wine when we have nothing but the taste
of physic in our mouths: the rich banquet tempts us not, when ‘our
very gorge rises’ within us: Love and Beauty fly from a bed twisted
into a thousand folds by restless lassitude and tormenting cares: the
nerve of pleasure is killed by the pains that shoot through the head or
rack the limbs: an indigestion seizes you with its leaden grasp and
giant force (down, Ambition!)—you shiver and tremble like a leaf in a
fit of the ague (Avarice, let go your palsied hold!). We then are in the
mood, without ghostly advice, to betake ourselves to the life of
‘hermit poor,
‘In pensive place obscure,’—
and should be glad to prevent the return of a fever raging in the

blood by feeding on pulse, and slaking our thirst at the limpid brook.
These sudden resolutions, however, or ‘vows made in pain as violent
and void,’ are generally of short duration; the excess and the sorrow
for it are alike selfish; and those repentances which are the most loud
and passionate are the surest to end speedily in a relapse; for both
originate in the same cause, the being engrossed by the prevailing
feeling (whatever it may be), and an utter incapacity to look beyond
it.
‘The Devil was sick, the Devil a monk would be:
The Devil grew well, the Devil a monk was he!’
It is amazing how little effect physical suffering or local

circumstances have upon the mind, except while we are subject to
their immediate influence. While the impression lasts, they are every
thing: when it is gone, they are nothing. We toss and tumble about in
a sick bed; we lie on our right side, we then change to the left; we
stretch ourselves on our backs, we turn on our faces; we wrap
ourselves up under the clothes to exclude the cold, we throw them off
to escape the heat and suffocation; we grasp the pillow in agony, we
fling ourselves out of bed, we walk up and down the room with hasty
or feeble steps; we return into bed; we are worn out with fatigue and
pain, yet can get no repose for the one, or intermission for the other;
we summon all our patience, or give vent to passion and petty rage:
nothing avails; we seem wedded to our disease, ‘like life and death in
disproportion met;’ we make new efforts, try new expedients, but
nothing appears to shake it off, or promise relief from our grim foe: it
infixes its sharp sting into us, or overpowers us by its sickly and
stunning weight: every moment is as much as we can bear, and yet
there seems no end of our lengthening tortures; we are ready to faint
with exhaustion, or work ourselves up to frenzy: we ‘trouble deaf
Heaven with our bootless prayers:’ we think our last hour is come, or
peevishly wish it were, to put an end to the scene; we ask questions
as to the origin of evil and the necessity of pain; we ‘moralise our
complaints into a thousand similes’; we deny the use of medicine in
toto, we have a full persuasion that all doctors are mad or knaves,
that our object is to gain relief, and theirs (out of the perversity of
human nature, or to seem wiser than we) to prevent it; we catechise
the apothecary, rail at the nurse, and cannot so much as conceive the
possibility that this state of things should not last for ever; we are
even angry at those who would give us encouragement, as if they
would make dupes or children of us; we might seek a release by
poison, a halter, or the sword, but we have not strength of mind
enough—our nerves are too shaken—to attempt even this poor
revenge—when lo! a change comes, the spell falls off, and the next
moment we forget all that has happened to us. No sooner does our
disorder turn its back upon us than we laugh at it. The state we have
been in, sounds like a dream, a fable; health is the order of the day,
strength is ours de jure and de facto; and we discard all uncalled-for
evidence to the contrary with a smile of contemptuous incredulity,
just as we throw our physic-bottles out of the window! I see (as I
awake from a short, uneasy doze) a golden light shine through my
white window-curtains on the opposite wall:—is it the dawn of a new
day, or the departing light of evening? I do not well know, for the
opium ‘they have drugged my posset with’ has made strange havoc
with my brain, and I am uncertain whether time has stood still, or
advanced, or gone backward. By ‘puzzling o’er the doubt,’ my
attention is drawn a little out of myself to external objects; and I
consider whether it would not administer some relief to my
monotonous languour, if I could call up a vivid picture of an evening
sky I witnessed a short while before, the white fleecy clouds, the
azure vault, the verdant fields and balmy air. In vain! The wings of
fancy refuse to mount from my bed-side. The air without has nothing
in common with the closeness within: the clouds disappear, the sky
is instantly overcast and black. I walk out in this scene soon after I
recover; and with those favourite and well-known objects interposed,
can no longer recall the tumbled pillow, the juleps or the labels, or
the unwholesome dungeon in which I was before immured. What is
contrary to our present sensations or settled habits, amalgamates
indifferently with our belief: the imagination rules over imaginary
themes, the senses and custom have a narrower sway, and admit but
one guest at a time. It is hardly to be wondered at that we dread
physical calamities so little beforehand: we think no more of them
the moment after they have happened. Out of sight, out of mind.
This will perhaps explain why all actual punishment has so little
effect; it is a state contrary to nature, alien to the will. If it does not
touch honour and conscience (and where these are not, how can it
touch them?) it goes for nothing: and where these are, it rather sears
and hardens them. The gyves, the cell, the meagre fare, the hard
labour are abhorrent to the mind of the culprit on whom they are
imposed, who carries the love of liberty or indulgence to
licentiousness; and who throws the thought of them behind him (the
moment he can evade the penalty,) with scorn and laughter,
‘Like Samson his green wythes.’[25]
So, in travelling, we often meet with great fatigue and

inconvenience from heat or cold, or rather accidents, and resolve
never to go a journey again; but we are ready to set off on a new
excursion to-morrow. We remember the landscape, the change of
scene, the romantic expectation, and think no more of the heat, the
noise, and dust. The body forgets its grievances, till they recur; but
imagination, passion, pride, have a longer memory and quicker
apprehensions. To the first the pleasure or the pain is nothing when
once over; to the last it is only then that they begin to exist. The line
in Metastasio,
‘The worst of every evil is the fear,’
is true only when applied to this latter sort.—It is curious that, on

coming out of a sick room, where one has been pent some time, and
grown weak and nervous, and looking at Nature for the first time, the
objects that present themselves have a very questionable and spectral
appearance, the people in the street resemble flies crawling about,
and seem scarce half-alive. It is we who are just risen from a torpid
and unwholesome state, and who impart our imperfect feelings of
existence, health, and motion to others. Or it may be that the
violence and exertion of the pain we have gone through make
common every-day objects seem unreal and unsubstantial. It is not
till we have established ourselves in form in the sitting-room,
wheeled round the arm-chair to the fire (for this makes part of our
re-introduction to the ordinary modes of being in all seasons,) felt
our appetite return, and taken up a book, that we can be considered
as at all restored to ourselves. And even then our first sensations are
rather empirical than positive; as after sleep we stretch out our
hands to know whether we are awake. This is the time for reading.
Books are then indeed ‘a world, both pure and good,’ into which we
enter with all our hearts, after our revival from illness and respite
from the tomb, as with the freshness and novelty of youth. They are
not merely acceptable as without too much exertion they pass the
time and relieve ennui; but from a certain suspension and deadening
of the passions, and abstraction from worldly pursuits, they may be
said to bring back and be friendly to the guileless and enthusiastic
tone of feeling with which we formerly read them. Sickness has
weaned us pro tempore from contest and cabal; and we are fain to be
docile and children again. All strong changes in our present pursuits
throw us back upon the past. This is the shortest and most complete
emancipation from our late discomfiture. We wonder that any one
who has read The History of a Foundling should labour under an
indigestion; nor do we comprehend how a perusal of the Faery
Queen should not ensure the true believer an uninterrupted
succession of halcyon days. Present objects bear a retrospective
meaning, and point to ‘a foregone conclusion.’ Returning back to life
with half-strung nerves and shattered strength, we seem as when we
first entered it with uncertain purposes and faltering aims. The
machine has received a shock, and it moves on more tremulously
than before, and not all at once in the beaten track. Startled at the
approach of death, we are willing to get as far from it as we can by
making a proxy of our former selves; and finding the precarious
tenure by which we hold existence, and its last sands running out, we
gather up and make the most of the fragments that memory has
stored up for us. Every thing is seen through a medium of reflection
and contrast. We hear the sound of merry voices in the street; and
this carries us back to the recollections of some country-town or
village-group—
‘We see the children sporting on the shore,
And hear the mighty waters roaring evermore.’
A cricket chirps on the hearth, and we are reminded of Christmas

gambols long ago. The very cries in the street seem to be of a former
date; and the dry toast eats very much as it did—twenty years ago. A
rose smells doubly sweet, after being stifled with tinctures and
essences; and we enjoy the idea of a journey and an inn the more for
having been bed-rid. But a book is the secret and sure charm to bring
all these implied associations to a focus. I should prefer an old one,
Mr. Lamb’s favourite, the Journey to Lisbon; or the Decameron, if I
could get it; but if a new one, let it be Paul Clifford. That book has
the singular advantage of being written by a gentleman, and not
about his own class. The characters he commemorates are every
moment at fault between life and death, hunger and a forced loan on
the public; and therefore the interest they take in themselves, and
which we take in them, has no cant or affectation in it, but is ‘lively,
audible, and full of vent.’ A set of well-dressed gentlemen picking
their teeth with a graceful air after dinner, endeavouring to keep
their cravats from the slightest discomposure, and saying the most
insipid things in the most insipid manner, do not make a scene. Well,
then, I have got the new paraphrase on the Beggar’s Opera, am fairly
embarked on it; and at the end of the first volume, where I am
galloping across the heath with the three highwaymen, while the
moon is shining full upon them, feel my nerves so braced, and my
spirits so exhilarated, that, to say truth, I am scarce sorry for the
occasion that has thrown me upon the work and the author—have
quite forgot my Sick Room, and am more than half ready to recant
the doctrine that a Free Admission to the theatre is
—‘The true pathos and sublime
Of human life’:—
for I feel as I read that if the stage shows us the masks of men and
the pageant of the world, books let us into their souls and lay open to
us the secrets of our own. They are the first and last, the most home-
felt, the most heartfelt of all our enjoyments.
FOOTMEN
The New Monthly Magazine.]

[September, 1830.
Footmen are no part of Christianity; but they are a very necessary

appendage to our happy Constitution in Church and State. What
would the bishop’s mitre be without these grave supporters to his
dignity? Even the plain presbyter does not dispense with his decent
serving-man to stand behind his chair and load his duly emptied
plate with beef and pudding, at which the genius of Ude turns pale.
What would become of the coronet-coach filled with elegant and
languid forms, if it were not for the triple row of powdered, laced,
and liveried footmen, clustering, fluttering, and lounging behind it?
What an idea do we not conceive of the fashionable belle who is
making the most of her time and tumbling over silks and satins
within at Sewell and Cross’s, or at the Bazaar in Soho-square, from
the tall lacquey in blue and silver with gold-headed cane, cocked-hat,
white thread stockings and large calves to his legs, who stands as her
representative without! The sleek shopman appears at the door, at an
understood signal the livery-servant starts from his position, the
coach-door flies open, the steps are let down, the young lady enters
the carriage as young ladies are taught to step into carriages, the
footman closes the door, mounts behind, and the glossy vehicle rolls
off, bearing its lovely burden and her gaudy attendant from the gaze
of the gaping crowd! Is there not a spell in beauty, a charm in rank
and fashion, that one would almost wish to be this fellow—to obey its
nod, to watch its looks, to breathe but by its permission, and to live
but for its use, its scorn, or pride?
Footmen are in general looked upon as a sort of supernumeraries
in society—they have no place assigned them in any Scotch
Encyclopædia—they do not come under any of the heads in Mr. Mill’s
Elements, or Mr. Maculloch’s Principles of Political Economy; and
they nowhere have had impartial justice done them, except in Lady
Booby’s love for one of that order. But if not ‘the Corinthian capitals
of polished society,’ they are ‘a graceful ornament to the civil order.’
Lords and ladies could not do without them. Nothing exists in this
world but by contrast. A foil is necessary to make the plainest truths
self-evident. It is the very insignificance, the nonentity as it were of
the gentlemen of the cloth, that constitutes their importance, and
makes them an indispensable feature in the social system, by setting
off the pretensions of their superiors to the best advantage. What
would be the good of having a will of our own, if we had not others
about us who are deprived of all will of their own, and who wear a
badge to say ‘I serve?’ How can we show that we are the lords of the
creation but by reducing others to the condition of machines, who
never move but at the beck of our caprices? Is not the plain suit of
the master wonderfully relieved by the borrowed trappings and
mock-finery of his servant? You see that man on horseback who
keeps at some distance behind another, who follows him as his
shadow, turns as he turns, and as he passes or speaks to him, lifts his
hand to his hat and observes the most profound attention—what is
the difference between these two men? The one is as well mounted,
as well fed, is younger and seemingly in better health than the other;
but between these two there are perhaps seven or eight classes of
society, each of whom is dependent on and trembles at the frown of
the other—it is a nobleman and his lacquey. Let any one take a stroll
towards the West-end of the town, South Audley or Upper
Grosvenor-street; it is then he will feel himself first entering into the
beau-ideal of civilized life, a society composed entirely of lords and
footmen! Deliver me from the filth and cellars of St. Giles’s, from the
shops of Holborn and the Strand, from all that appertains to middle
and to low life; and commend me to the streets with the straw at the
doors and hatchments overhead to tell us of those who are just born
or who are just dead, and with groups of footmen lounging on the
steps and insulting the passengers—it is then I feel the true dignity
and imaginary pretensions of human nature realised! There is here
none of the squalidness of poverty, none of the hardships of daily
labour, none of the anxiety and petty artifice of trade; life’s business
is changed into a romance, a summer’s dream, and nothing painful,
disgusting, or vulgar intrudes. All is on a liberal and handsome scale.
The true ends and benefits of society are here enjoyed and
bountifully lavished, and all the trouble and misery banished, and
not even allowed so much as to exist in thought. Those who would
find the real Utopia, should look for it somewhere about Park-lane or
May Fair. It is there only any feasible approach to equality is made—
for it is like master like man. Here, as I look down Curzon Street, or
catch a glimpse of the taper spire of South Audley Chapel, or the
family-arms on the gate of Chesterfield-House, the vista of years
opens to me, and I recall the period of the triumph of Mr. Burke’s
‘Reflections on the French Revolution,’ and the overthrow of ‘The
Rights of Man!’ You do not indeed penetrate to the interior of the
mansion where sits the stately possessor, luxurious and refined; but
you draw your inference from the lazy, pampered, motley crew
poured forth from his portals. This mealy-coated, moth-like,
butterfly-generation, seem to have no earthly business but to enjoy
themselves. Their green liveries accord with the budding leaves and
spreading branches of the trees in Hyde Park—they seem ‘like
brothers of the groves’—their red faces and powdered heads
harmonise with the blossoms of the neighbouring almond-trees, that
shoot their sprays over old-fashioned brick-walls. They come forth
like grasshoppers in June, as numerous and as noisy. They bask in
the sun and laugh in your face. Not only does the master enjoy an
uninterrupted leisure and tranquillity—those in his employment
have nothing to do. He wants drones, not drudges, about him, to
share his superfluity, and give a haughty pledge of his exemption
from care. They grow sleek and wanton, saucy and supple. From
being independent of the world, they acquire the look of gentlemen’s
gentlemen. There is a cast of the aristocracy, with a slight shade of
distinction. The saying, ‘Tell me your company, and I’ll tell you your
manners,’ may be applied cum grano salis to the servants in great
families. Mr. N—— knew an old butler who had lived with a
nobleman so long, and had learned to imitate his walk, look, and way
of speaking, so exactly that it was next to impossible to tell them
apart. See the porter in the great leather-chair in the hall—how big,
and burly, and self-important he looks; while my Lord’s gentleman
(the politician of the family) is reading the second edition of ‘The
Courier’ (once more in request) at the side window, and the footman
is romping, or taking tea with the maids in the kitchen below. A
match-girl meanwhile plies her shrill trade at the railing; or a gipsey-
woman passes with her rustic wares through the street, avoiding the
closer haunts of the city. What a pleasant farce is that of ‘High Life
Below Stairs!’ What a careless life do the domestics of the Great lead!
For, not to speak of the reflected self-importance of their masters
and mistresses, and the contempt with which they look down on the
herd of mankind, they have only to eat and drink their fill, talk the
scandal of the neighbourhood, laugh at the follies, or assist the
intrigues of their betters, till they themselves fall in love, marry, set
up a public house, (the only thing they are fit for,) and without habits
of industry, resources in themselves, or self-respect, and drawing
fruitless comparisons with the past, are, of all people, the most
miserable! Service is no inheritance; and when it fails, there is not a
more helpless, or more worthless set of devils in the world. Mr. C——
used to say he should like to be a footman to some elderly lady of
quality, to carry her prayer-book to church, and place her cassock
right for her. There can be no doubt that this would have been better,
and quite as useful as the life he has led, dancing attendance on
Prejudice, but flirting with Paradox in such a way as to cut himself
out of the old lady’s will. For my part, if I had to choose, I should
prefer the service of a young mistress, and might share the fate of the
footman recorded in heroic verse by Lady Wortley Montagu.
Certainly it can be no hard duty, though a sort of forlorn hope, to
have to follow three sisters, or youthful friends, (resembling the
three Graces,) at a slow pace, and with grave demeanour, from
Cumberland Gate to Kensington Gardens—to be there shut out, a
privation enhancing the privilege, and making the sense of distant,
respectful, idolatrous admiration more intense—and then, after a
brief interval lost in idle chat, or idler reverie, to have to follow them
back again, observing, not observed, to keep within call, to watch
every gesture, to see the breeze play with the light tresses or lift the
morning robe aside, to catch the half-suppressed laugh, and hear the
low murmur of indistinct words and wishes, like the music of the
spheres. An amateur footman would seem a more rational
occupation than that of an amateur author, or an amateur artist. An
insurmountable barrier, if it excludes passion, does not banish
sentiment, but draws an atmosphere of superstitious, trembling
apprehension round the object of so much attention and respect;
nothing makes women seem so much like angels as always to see,
never to converse with them; and those whom we have to dangle a
cane after must, to a lacquey of any spirit, appear worthy to wield
sceptres.
But of all situations of this kind, the most enviable is that of a
lady’s maid in a family travelling abroad. In the obtuseness of
foreigners to the nice gradations of English refinement and manners,
the maid has not seldom a chance of being taken for the mistress—a
circumstance never to be forgot! See our Abigail mounted in the
dicky with my Lord, or John, snug and comfortable—setting out on
the grand tour as fast as four horses can carry her, whirled over the
‘vine-covered hills and gay regions of France,’ crossing the Alps and
Apennines in breathless terror and wonder—frightened at a
precipice, laughing at her escape—coming to the inn, going into the
kitchen to see what is to be had—not speaking a word of the
language, except what she picks up, ‘as pigeons pick up peas:‘—the
bill paid, the passport visé, the horses put to, and au route again—
seeing every thing, and understanding nothing, in a full tide of
health, fresh air, and animal spirits, and without one qualm of taste
or sentiment, and arriving at Florence, the city of palaces, with its
amphitheatre of hills and olives, without suspecting that such a
person as Boccacio, Dante, or Galileo, had ever lived there, while her
young mistress is puzzled with the varieties of the Tuscan dialect, is
disappointed in the Arno, and cannot tell what to make of the statue
of David by Michael Angelo, in the Great Square. The difference is,
that the young lady, on her return, has something to think of; but the
maid absolutely forgets every thing, and is only giddy and out of
breath, as if she had been up in a balloon.
‘No more: where ignorance is bliss,
’Tis folly to be wise!’
English servants abroad, notwithstanding the comforts they enjoy,

and although travelling as it were en famille, must be struck with the
ease and familiar footing on which foreigners live with their
domestics, compared with the distance and reserve with which they
are treated. The housemaid (la bonne) sits down in the room, or
walks abreast with you in the street; and the valet who waits behind
his master’s chair at table, gives Monsieur his advice or opinion
without being asked for it. We need not wonder at this familiarity
and freedom, when we consider that those who allowed it could
(formerly at least, when the custom began) send those who
transgressed but in the smallest degree to the Bastille or the galleys
at their pleasure. The licence was attended with perfect impunity.
With us the law leaves less to discretion; and by interposing a real
independence (and plea of right) between the servant and master,
does away with the appearance of it on the surface of manners. The
insolence and tyranny of the Aristocracy fell more on the
tradespeople and mechanics than on their domestics, who were
attached to them by a semblance of feudal ties. Thus an upstart lady
of quality (an imitator of the old school) would not deign to speak to
a milliner while fitting on her dress, but gave her orders to her
waiting-women to tell her what to do. Can we wonder at twenty
reigns of terror to efface such a feeling?
I have alluded to the inclination in servants in great houses to ape
the manners of their superiors, and to their sometimes succeeding.
What facilitates the metamorphosis is, that the Great, in their
character of courtiers, are a sort of footmen in their turn. There is
the same crouching to interest and authority in either case, with the
same surrender or absence of personal dignity—the same submission
to the trammels of outward form, with the same suppression of
inward impulses—the same degrading finery, the same pretended
deference in the eye of the world, and the same lurking contempt
from being admitted behind the scenes, the same heartlessness, and
the same eye-service—in a word, they are alike puppets governed by
motives not their own, machines made of coarser or finer materials.
It is not, therefore, surprising, if the most finished courtier of the day
cannot, by a vulgar eye, be distinguished from a gentleman’s servant.
M. de Bausset, in his amusing and excellent Memoirs, makes it an
argument of the legitimacy of Napoleon’s authority, that from
denying it, it would follow that his lords of the bed-chamber were
valets, and he himself (as prefect of the palace) no better than head-
cook. The inference is logical enough. According to the author’s view,
there was no other difference between the retainers of the court and
the kitchen than the rank of the master!
I remember hearing it said that ‘all men were equal but footmen.’
But of all footmen the lowest class is literary footmen. These consist
of persons who, without a single grain of knowledge, taste, or feeling,
put on the livery of learning, mimic its phrases by rote, and are
retained in its service by dint of quackery and assurance alone. As
they have none of the essence, they have all the externals of men of
gravity and wisdom. They wear green spectacles, walk with a peculiar
strut, thrust themselves into the acquaintance of persons they hear
talked of, get introduced into the clubs, are seen reading books they
do not understand at the Museum and public libraries, dine (if they
can) with lords or officers of the Guards, abuse any party as low to
show what fine gentlemen they are, and the next week join the same
party to raise their own credit and gain a little consequence, give
themselves out as wits, critics, and philosophers (and as they have
never done any thing, no man can contradict them), and have a great
knack of turning editors, and not paying their contributors. If you get
five pounds from one of them, he never forgives it. With the proceeds
thus appropriated, the book-worm graduates a dandy, hires
expensive apartments, sports a tandem, and it is inferred that he
must be a great author who can support such an appearance with his
pen, and a great genius who can conduct so many learned works
while his time is devoted to the gay, the fair, and the rich. This
introduces him to new editorships, to new and more select
friendships, and to more frequent and importunate demands from
debts and duns. At length the bubble bursts and disappears, and you
hear no more of our classical adventurer, except from the invectives
and self-reproaches of those who took him for a great scholar from
his wearing green spectacles and Wellington-boots. Such a candidate
for literary honours bears the same relation to the man of letters,
that the valet with his second-hand finery and servile airs does to his
master.
ON THE WANT OF MONEY
The Monthly Magazine.]

[January, 1827.
It is hard to be without money. To get on without it is like

travelling in a foreign country without a passport—you are stopped,
suspected, and made ridiculous at every turn, besides being
subjected to the most serious inconveniences. The want of money I
here allude to is not altogether that which arises from absolute
poverty—where there is a downright absence of the common
necessaries of life, this must be remedied by incessant hard labour,
and the least we can receive in return is a supply of our daily wants—
but that uncertain, casual, precarious mode of existence, in which the
temptation to spend remains after the means are exhausted, the
want of money joined with the hope and possibility of getting it, the
intermediate state of difficulty and suspense between the last guinea
or shilling and the next that we may have the good luck to encounter.
This gap, this unwelcome interval constantly recurring, however
shabbily got over, is really full of many anxieties, misgivings,
mortifications, meannesses, and deplorable embarrassments of every
description. I may attempt (this essay is not a fanciful speculation) to
enlarge upon a few of them.
It is hard to go without one’s dinner through sheer distress, but
harder still to go without one’s breakfast. Upon the strength of that
first and aboriginal meal, one may muster courage to face the
difficulties before one, and to dare the worst: but to be roused out of
one’s warm bed, and perhaps a profound oblivion of care, with
golden dreams (for poverty does not prevent golden dreams), and
told there is nothing for breakfast, is cold comfort for which one’s
half-strung nerves are not prepared, and throws a damp upon the
prospects of the day. It is a bad beginning. A man without a breakfast
is a poor creature, unfit to go in search of one, to meet the frown of
the world, or to borrow a shilling of a friend. He may beg at the
corner of a street—nothing is too mean for the tone of his feelings—
robbing on the highway is out of the question, as requiring too much
courage, and some opinion of a man’s self. It is, indeed, as old Fuller,
or some worthy of that age, expresses it, ‘the heaviest stone which
melancholy can throw at a man,’ to learn, the first thing after he rises
in the morning, or even to be dunned with it in bed, that there is no
loaf, tea, or butter in the house, and that the baker, the grocer, and
butterman have refused to give any farther credit. This is taking one
sadly at a disadvantage. It is striking at one’s spirit and resolution in
their very source,—the stomach—it is attacking one on the side of
hunger and mortification at once; it is casting one into the very mire
of humility and Slough of Despond. The worst is, to know what face
to put upon the matter, what excuse to make to the servants, what
answer to send to the tradespeople; whether to laugh it off, or be
grave, or angry, or indifferent; in short, to know how to parry off an
evil which you cannot help. What a luxury, what a God’s-send in such
a dilemma, to find a half-crown which had slipped through a hole in
the lining of your waistcoat, a crumpled bank-note in your breeches-
pocket, or a guinea clinking in the bottom of your trunk, which had
been thoughtlessly left there out of a former heap! Vain hope!
Unfounded illusion! The experienced in such matters know better,
and laugh in their sleeves at so improbable a suggestion. Not a
corner, not a cranny, not a pocket, not a drawer has been left
unrummaged, or has not been subjected over and over again to more
than the strictness of a custom-house scrutiny. Not the slightest
rustle of a piece of bank-paper, not the gentlest pressure of a piece of
hard metal, but would have given notice of its hiding-place with
electrical rapidity, long before, in such circumstances. All the variety
of pecuniary resources which form a legal tender on the current coin
of the realm, are assuredly drained, exhausted to the last farthing
before this time. But is there nothing in the house that one can turn
to account! Is there not an old family-watch, or piece of plate, or a
ring, or some worthless trinket that one could part with? nothing
belonging to one’s-self or a friend, that one could raise the wind
upon, till something better turns up? At this moment an old clothes-
man passes, and his deep, harsh tones sound like an intended insult
on one’s distress, and banish the thought of applying for his
assistance, as one’s eye glanced furtively at an old hat or a great coat,
hung up behind a closet-door. Humiliating contemplations!
Miserable uncertainty! One hesitates, and the opportunity is gone by;
for without one’s breakfast, one has not the resolution to do any
thing!—The late Mr. Sheridan was often reduced to this unpleasant
predicament. Possibly he had little appetite for breakfast himself; but
the servants complained bitterly on this head, and said that Mrs.
Sheridan was sometimes kept waiting for a couple of hours, while
they had to hunt through the neighbourhood, and beat up for coffee,
eggs, and French rolls. The same perplexity in this instance appears
to have extended to the providing for the dinner; for so sharp-set
were they, that to cut short a debate with a butcher’s apprentice
about leaving a leg of mutton without the money, the cook clapped it
into the pot: the butcher’s boy, probably used to such encounters,
with equal coolness took it out again, and marched off with it in his
tray in triumph. It required a man to be the author of The School
for Scandal, to run the gauntlet of such disagreeable occurrences
every hour of the day. There was one comfort, however, that poor
Sheridan had: he did not foresee that Mr. Moore would write his
Life![26]
The going without a dinner is another of the miseries of wanting
money, though one can bear up against this calamity better than the
former, which really ‘blights the tender blossom and promise of the
day.’ With one good meal, one may hold a parley with hunger and
moralize upon temperance. One has time to turn one’s-self and look
about one—to ‘screw one’s courage to the sticking-place,’ to graduate
the scale of disappointment, and stave off appetite till supper-time.
You gain time, and time in this weather-cock world is everything.
You may dine at two, or at six, or seven—as most convenient. You
may in the meanwhile receive an invitation to dinner, or some one
(not knowing how you are circumstanced) may send you a present of
a haunch of venison or a brace of pheasants from the country, or a
distant relation may die and leave you a legacy, or a patron may call
and overwhelm you with his smiles and bounty,
‘As kind as kings upon their coronation-day;’
or there is no saying what may happen. One may wait for dinner—
breakfast admits of no delay, of no interval interposed between that
and our first waking thoughts.[27] Besides, there are shifts and
devices, shabby and mortifying enough, but still available in case of
need. How many expedients are there in this great city (London),
time out of mind and times without number, resorted to by the
dilapidated and thrifty speculator, to get through this grand difficulty
without utter failure! One may dive into a cellar, and dine on boiled
beef and carrots for tenpence, with the knives and forks chained to
the table, and jostled by greasy elbows that seem to make such a
precaution not unnecessary (hunger is proof against indignity!)—or
one may contrive to part with a superfluous article of wearing
apparel, and carry home a mutton-chop and cook it in a garret; or
one may drop in at a friend’s at the dinner-hour, and be asked to stay
or not; or one may walk out and take a turn in the Park, about the
time, and return home to tea, so as at least to avoid the sting of the
evil—the appearance of not having dined. You then have the laugh on
your side, having deceived the gossips, and can submit to the want of
a sumptuous repast without murmuring, having saved your pride,
and made a virtue of necessity. I say all this may be done by a man
without a family (for what business has a man without money with
one?—See English Malthus and Scotch Macculloch)—and it is only
my intention here to bring forward such instances of the want of
money as are tolerable both in theory and practice. I once lived on
coffee (as an experiment) for a fortnight together, while I was
finishing the copy of a half-length portrait of a Manchester
manufacturer, who had died worth a plum. I rather slurred over the
coat, which was a reddish brown, ‘of formal cut,’ to receive my five
guineas, with which I went to market myself, and dined on sausages
and mashed potatoes, and while they were getting ready, and I could
hear them hissing in the pan, read a volume of Gil Blas, containing
the account of the fair Aurora. This was in the days of my youth.
Gentle reader, do not smile! Neither Monsieur de Very, nor Louis
XVIII., over an oyster-pâté, nor Apicius himself, ever understood the
meaning of the word luxury, better than I did at that moment! If the
want of money has its drawbacks and disadvantages, it is not without
its contrasts and counterbalancing effects, for which I fear nothing
else can make us amends. Amelia’s hashed mutton is immortal; and
there is something amusing, though carried to excess and caricature
(which is very unusual with the author) in the contrivances of old
Caleb, in ‘The Bride of Lammermuir,’ for raising the wind at
breakfast, dinner, and supper-time. I recollect a ludicrous instance of
a disappointment in a dinner which happened to a person of my
acquaintance some years ago. He was not only poor but a very poor
creature, as will be imagined. His wife had laid by fourpence (their
whole remaining stock) to pay for the baking of a shoulder of mutton
and potatoes, which they had in the house, and on her return home
from some errand, she found he had expended it in purchasing a new
string for a guitar. On this occasion a witty friend quoted the lines
from Milton:
‘And ever against eating cares,
Wrap me in soft Lydian airs!’
Defoe, in his Life of Colonel Jack, gives a striking picture of his

young beggarly hero sitting with his companion for the first time in
his life at a three-penny ordinary, and the delight with which he
relished the hot smoking soup, and the airs with which he called
about him—‘and every time,’ he says, ‘we called for bread, or beer, or
whatever it might be, the waiter answered, “coming, gentlemen,
coming;” and this delighted me more than all the rest!’ It was about
this time, as the same pithy author expresses it, ‘the Colonel took
upon him to wear a shirt!’ Nothing can be finer than the whole of the
feeling conveyed in the commencement of this novel, about wealth
and finery from the immediate contrast of privation and poverty.
One would think it a labour, like the Tower of Babel, to build up a
beau and a fine gentleman about town. The little vagabond’s
admiration of the old man at the banking-house, who sits
surrounded by heaps of gold as if it were a dream or poetic vision,
and his own eager anxious visits, day by day, to the hoard he had
deposited in the hollow tree, are in the very foremost style of truth
and nature. See the same intense feeling expressed in Luke’s address
to his riches in the City Madam, and in the extraordinary raptures of
the ‘Spanish Rogue’ in contemplating and hugging his ingots of pure
gold and Spanish pieces of eight: to which Mr. Lamb has referred in
excuse for the rhapsodies of some of our elder poets on this subject,
which to our present more refined and tamer apprehensions sound
like blasphemy.[28] In earlier times, before the diffusion of luxury, of
knowledge, and other sources of enjoyment had become common,
and acted as a diversion to the cravings of avarice, the passionate
admiration, the idolatry, the hunger and thirst of wealth and all its
precious symbols, was a kind of madness or hallucination, and
Mammon was truly worshipped as a god!
It is among the miseries of the want of money, not to be able to pay
your reckoning at an inn—or, if you have just enough to do that, to
have nothing left for the waiter;—to be stopped at a turnpike gate,
and forced to turn back;—not to venture to call a hackney-coach in a
shower of rain—(when you have only one shilling left yourself, it is a
bore to have it taken out of your pocket by a friend, who comes into
your house eating peaches in a hot summer’s day, and desiring you to
pay for the coach in which he visits you);—not to be able to purchase
a lottery-ticket, by which you might make your fortune, and get out
of all your difficulties;—or to find a letter lying for you at a country
post-office, and not to have money in your pocket to free it, and be
obliged to return for it the next day. The letter so unseasonably
withheld may be supposed to contain money, and in this case there is
a foretaste, a sort of actual possession taken through the thin folds of
the paper and the wax, which in some measure indemnifies us for the
delay: the bank-note, the post-bill seems to smile upon us, and shake
hands through its prison bars;—or it may be a love-letter, and then
the tantalization is at its height: to be deprived in this manner of the
only consolation that can make us amends for the want of money, by
this very want—to fancy you can see the name—to try to get a peep at
the hand-writing—to touch the seal, and yet not dare to break it open
—is provoking indeed—the climax of amorous and gentlemanly
distress. Players are sometimes reduced to great extremity, by the
seizure of their scenes and dresses, or (what is called) the property of
the theatre, which hinders them from acting; as authors are
prevented from finishing a work, for want of money to buy the books
necessary to be consulted on some material point or circumstance, in
the progress of it. There is a set of poor devils, who live upon a
printed prospectus of a work that never will be written, for which
they solicit your name and half-a-crown. Decayed actresses take an
annual benefit at one of the theatres; there are patriots who live upon
periodical subscriptions, and critics who go about the country
lecturing on poetry. I confess I envy none of these; but there are
persons who, provided they can live, care not how they live—who are
fond of display, even when it implies exposure; who court notoriety

Practical Applications of Computational Biology and Bioinformatics 16th International Conference PACBB 2022 Florentino Fdez-Riverola

Uploaded by

Copyright:

Available Formats

Practical Applications of Computational Biology and Bioinformatics 16th International Conference PACBB 2022 Florentino Fdez-Riverola

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Practical Applications of Computational Biology and Bioinformatics 16th International Conference PACBB 2022 Florentino Fdez-Riverola

Uploaded by

Copyright:

Available Formats

Full download test bank at ebookmeta.

Practical Applications of Computational Biology

Download More ebooks from https://ebookmeta.com

GeNeDis 2022 Computational Biology and Bioinformatics

Bioinformatics and Computational Biology 1st Edition

Parallel Computational Technologies 16th International

Bioinformatics and Computational Biology A Primer for

Bioinformatics: A Practical Guide to Next Generation

Enterprise, Business-Process and Information Systems

Algorithmic Aspects in Information and Management: 16th

Advances on Broad Band Wireless Computing Communication

For proposals from Asia please contact Aninda Bose (aninda.bose@springer.com).

Mohd Saberi Mohamad Simona Caraiman

Ana Belén Gil-González

ISSN 2367-3370 ISSN 2367-3389 (electronic)

The success of bioinformatics in recent years has been prompted by research in

Vigo, Spain Florentino Fdez-Riverola

Program Committee Chairs

Organising Committee Chairs

Florentino Fdez-Riverola, University of Vigo, Spain

Grabriella Panuccio, Istituto Italiano di Tecnología, Italy

Local Organizing Committee

Pierpaolo Vittorini (Co-chair), University of L’aquila, Italy

Juan M. Corchado Rodríguez, University of Salamanca, Spain; AIR Institute, Spain

Vera Afreixo, University of Aveiro, Portugal

Rosalba Giugno, University of Verona, Italy

TooT-BERT-T: A BERT Approach on Discriminating Transport

Exploring Xylella fastidiosa’s Metabolic Traits Using a GSM

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Hamed Ghazikhani and Gregory Butler

Abstract Transmembrane transport proteins (transporters) serve a crucial role for

Keywords Transmembrane transport proteins · Machine learning · BERT ·

H. Ghazikhani (B) · G. Butler

massive datasets of naturally occurring proteins with enough information to build

Aplop and Butler [4, 5] provide a comprehensive overview of transport protein

3 Materials and Methods

Table 1 DS-T: transport proteins dataset

3.2 Protein Sequence Representation

As multiple studies demonstrate, representation learning, a branch of machine learn-

Multi H ead(Q, K , V ) = Concat (head1 , ..., headn )W o (1)

3.3 Fine-Tuning a BERT Model

Table 2 Fine-tuning ProtBERT-BFD and MembraneBERT hyperparameters

3.4 Logistic Regression

Logistic Regression is a widely used classification technique in medical/biological

3.6 Evaluation Metrics

MCC is an acronym for Matthew’s Correlation Coefficient. For imbalanced data,

4 Results and Discussion

4.1 Fine-Tuning ProtBERT-BFD and MembraneBERT

We compared both representations of ProtBERT-BFD and MembraneBERT, without

Table 3 Logistic Regression performance with ProtBERT-BFD and MembraneBERT

fine-tuned representations than MembraneBERT, with the exception of the frozen

4.2 Logistic Regression with Fine-Tuned ProtBERT-BFD

We selected Logistic Regression as a preliminary good binary classifier because it

both fine-tuned ProtBERT-BFD and MembraneBERT representations performs well,

4.3 Comparison of TooT-BERT-T with State-of-the-Art

Table 4 Comparative performance of TooT-BERT-T with state-of-the-art

Fig. 2 Comparison of methodologies

Fig. 3 TooT-BERT-T Predicted values

TooT-BERT-T distinguishes transport proteins from non-transport proteins using the