Protein contact map prediction using bi-directional recurrent neural network

TL;DR

This paper introduces a deep learning model using bi-directional recurrent neural networks for protein contact map prediction, achieving significantly higher accuracy and advancing the protein folding research.

Contribution

The study presents a novel neural network architecture that substantially improves contact map prediction accuracy over previous models.

Findings

Achieved 0.80 accuracy on Protein Data Bank data

Significantly outperforms previous models in contact prediction

Marks a major step forward in protein folding research

Abstract

Given native 2D contact map, protein 3D structure could be reconstructed with accuracy of 2A or better, and such reconstruction is a feasible computational approach for protein folding problem. The prediction accuracy from traditional methods is generally too poor to useful, but the recent deep learning model has significantly improved the accuracy. In this study, we proposed a neural network model comprising a bi-directional recurrent neural network and artificial neural network. Over the non-redundant database of all available protein 3D structures in Protein Data Bank, this deep learning model achieved an accuracy of 0.80, much higher than those of previous models. This study represents a major breakthrough in protein 2D contact map prediction and likely a major step forward for the protein folding problem.