fRNAdb is a database that helps in annotating non-coding transcripts acquired from publicly available databases. H-inv: human full-length non-coding cDNAs (1); NONCODE: experimentally validated non-coding transcripts (2); and RNAdb: non-coding transcripts curated from the literature, human chromosome 7 project, and RIKEN antisense pipeline and other putative non-coding RNAs (3). Details are shown in Table 1. Each transcript is analyzed for various features such as maximum ORF length, the number of protein homologs, the average conservation score, transcription regulatory element motifs, existence of CpG islands and so on (listed in Table 2) that help in filtering out promising non-coding candidates. Transcripts can be filtered with fRNAdb's main listing interface in many different ways (see Figure 1). This main listing interface is linked to our custom UCSC Genome Browser (4) for functional RNAs equipped with our RNA-specific original custom tracks that are specific to screening of functional RNA. Users can inspect a transcript of interest from a genomic view with rich genomic information surrounding the mapped transcript. The information includes the UCSC original tracks such as known genes, genome conservation and Affymetrix transcriptome tracks (5), and our original tracks such as conserved potential secondary structure, existence of known RNA secondary structure motifs and significant RNA secondary structure Z-score regions (for details see Table 3).

Open in a separate window

Figure 1

The first page shows a set of selection interfaces (A) and the listing table of 13 693 transcripts (B).

fRNAdb provides two types of interfaces. The first page presents a list of all transcripts rendered as a table with 35 columns including ones for the attributes described in Table 2 (Figure 1B). The tabular control panel is placed above the table, which presents five tabs labeled ‘Basic’, ‘DB/ID’, ‘Expert’, ‘Sort’ and ‘Column’ (Figure 1A). The Basic tab contains the basic filters: a collection of frequently used filters that provide simple and quick selection of transcripts that match common criteria of functional non-coding RNAs. For example, checking ‘Mapped’ to select only genome-mapped transcripts, ‘Well conserved at best (Max > 50%)’ for transcripts that have maximum conservation score >50% among 17 vertebrates (4) in their exonic regions, ‘EST-supported’ for reliable expression evidence, ‘Tiny ORF (<40 aa)’ enriching for non-coding transcripts, ‘Low Repeat Coverage (<30%)’ for no repeat element contamination, ‘No protein homolog’ for another condition which enriches non-coding transcripts, ‘No overlapping known gene’ is for removing the possibility of being part of a protein-coding gene transcript. After checking the boxes, the ‘refresh’ button runs filtering action and presents results. Our example conditions yield nine hits including one H-inv non-protein coding cDNA and eight RNAdb literature-curated miRNAs. In other words, these criteria match real functional RNAs and also indicate that one non-coding transcript shares the same properties. Clicking on the ID of this transcript produces a detailed view of this transcript shown in Figure 2. This feature visualizer shows graphical representation of a variety of sequence elements found in the transcript including cis-regulatory elements, repeat elements, EST mapping regions and six frame stop codon positions. There are many different ways to filter these non-coding transcripts and there are many more potential candidates hidden in this dataset. More details of the basic filters are provided on the website.

Open in a separate window

Figure 2

mRNA view of a transcript. Regulatory elements, EST positions, splice positions, repeat elements, six frame stop codons are visualized along with the full span of a cDNA.

The rest of the tabs offer additional functionality to further improve usability. The DB/ID tab contains DB selection and ID selection boxes. The DB selection box allows you to limit the target databases from currently available databases: H-inv, NONCODE and RNAdb. The ID selection box lets you choose target transcripts that match given string patterns. For example, specifying ‘FR000001’ (fRNAdb ID) in this box limits the target transcript FR000001 alone. The wild-card ‘%’ is allowed for pattern matching. Specifying ‘LIT%’ lets you limit the search to targets whose original IDs start with ‘LIT’. The string pattern is matched against ID, Acc. and Original columns. The Expert tab provides an interface to specify multiple conditions that let you perform more complex filtering than the basic filters. Please refer to the website for more details about the expert filters. The Sort tab has a sorting interface that lets you sort the table with multiple sorting keys. The Column tab allows you to limit visible columns of the main listing table. Since the 35-column table is too wide for ordinary browsers to display on a single screen, you can narrow the width of the table with this interface for better visibility.

We mirrored the UCSC Genome Browser and added our custom tracks specific to functional RNAs and miRNAs as shown in Tables 3 and ​and4.4. Most of the tracks have their own sources and reference papers. Our original tracks are RNA clusters, Rfam seed folds, tRNAscan-SE, Ultra Conserved Elements 17way and Z-score (details are shown in Table 3). Besides, we mapped RNA sequences from public functional RNA sequence databases including Erdmann (6), NONCODE, RNAdb and Rfam. The UCSC Genome Browser has several tracks for miRNA genes and targets but we added more tracks including miRBase (7) known miRNA genes, miRNAMap (8) and Berezikov's predicted miRNA genes (9), TarBase (10) known miRNA targets, and predicted miRNA targets from RNAhybrid (11), PicTar 4 species and 5 species (12), miRBase targets and T-ScanS miRNA targets (13). Our custom tracks can be downloaded by using Table browser which can be accessed via ‘Table’ menu of the UCSC Genome Browser.

In the near future, fRNAdb will include more transcripts from other sequence databases or non-coding gene prediction results. For example, Human Accelerated Region (14) is currently included as our custom track of the Genome Browser. Sequences of these non-coding gene candidates will be included in fRNAdb. We will also add more attributes to fRNAdb. Especially attributes representing expression patterns of the transcripts or protein genes related to the transcripts.