Improving protein function prediction with synthetic feature samples created by generative adversarial networks

Abstract

Protein function prediction is a challenging but important task in bioinformatics. Many prediction methods have been developed, but are still limited by the bottleneck on training sample quantity. Therefore, it is valuable to develop a data augmentation method that can generate high-quality synthetic samples to further improve the accuracy of prediction methods. In this work, we propose a novel generative adversarial networks-based method, namely FFPred-GAN, to accurately learn the high-dimensional distributions of protein sequence-based biophysical features and also generate high-quality synthetic protein feature samples. The experimental results suggest that the synthetic protein feature samples are successful in improving the prediction accuracy for all three domains of the Gene Ontology through augmentation of the original training protein feature samples. AbstractAuthor summary

Improving Protein Function Annotation via Unsupervised Pre-training: Robustness, Efficiency, and Insights

“…On these tasks, pretraining improved performance across multiple model architectures, though follow-up work suggests that similar results could be achieved without pretraining [50]. An additional line of work [33,61,62] benchmarks various protein sequence embeddings on the CAFA benchmark [64], though BERT and related architectures are not considered.…”

PFP-WGAN: Protein function prediction by discovering Gene Ontology term correlations with generative adversarial networks

“…Works in [37,38] use GANs to analyze gene expression profiles and works in [39,40] attempt to synthesize genes and promoters by GANs. Recently authors in [41] have been proposed to perform data augmentation to generate synthetic training samples by a GAN to improve a classifier accuracy for annotating proteins.…”