GAN-PCL: An Efficient Protein Subchloroplast Site Predictor with GAN-based Data Augmented and Feature Fusion

Chloroplasts are important for photosynthesis, and proteins are distributed in different chloroplast regions to perform different functions. Although many computational methods for protein subchloroplast localization have proposed, prediction accuracy is still limited due to scarce and severely unbalanced samples. The development of a model that generates high-quality samples to supplement existing data to improve prediction performance is great significance for the study of chloroplast abundance in early plant development and the regulation of photosynthetic efficiency in agricultural production. This paper proposes a protein subchloroplast site predictor GAN-PCL based on data augmentation and sequence feature fusion with an adversarial neural network GAN, which effectively solves the problem. The experimental results show that GAN-PCL has good prediction performance and generalization ability compared with current advanced predictors.