Protein Residue Contact Prediction Based on Deep Learning and Massive Statistical Features from Multi-Sequence Alignment

Sequence-based protein tertiary structure prediction is of fundamental importance because the function a ultimately depends on its 3D structure. An accurate residue-residue contact map one essential elements for current ab initio protocols prediction. Recently, with combination deep learning and direct coupling techniques, performance residue has achieved significant progress. However, considerable number Deep-Learning (DL)-based methods are usually time-consuming, mainly they rely different categories data types third-party programs. In this research, we transformed complex biological problem into pure computational through statistics artificial intelligence. We have accordingly proposed feature extraction method to obtain various statistical information from only multi-sequence alignment, followed by training DL model based massive information. The robust in terms test sets, showed high reliability confidence score, could efficiency achieve comparable precisions that relying multi-source inputs.