Protein residue contact diagram prediction method based on deep residual neural network

Abstract

The invention discloses a protein residue contact diagram prediction method based on a deep residual neural network. The method comprises the following steps: firstly, inputting a protein sequence tobe subjected to protein residue contact diagram prediction; then, converting the protein sequence into a 20 * L matrix by utilizing one-hot representation forms of 20 common amino acids for the protein sequence, digitizing the protein sequence information, and calculating by utilizing the 20 * L matrix to obtain a 20 * L * L covariance tensor, i.e., the characteristics of the input network; secondly, establishing a deep residual neural network framework, collecting protein sequences and tags of existing protein contact tags from a PDB library, calculating feature tensors of the protein sequences, forming a data set by the feature tensors and the corresponding tags, and learning a prediction model on the data set by using the deep residual neural network framework; and finally, inputting the protein characteristic tensor to be subjected to protein residue contact diagram prediction into the model to obtain a protein sequence residue contact diagram. The method is low in calculation costand high in prediction precision.

Description

technical field [0001] The invention belongs to the fields of bioinformatics and computer applications, and in particular relates to a method for predicting protein residue contact maps based on deep residual neural networks. Background technique [0002] Protein is one of the most important macromolecules in biology and has a wide range of functions. The interaction between protein molecules is realized through the contact interaction between some residues. This interaction is ubiquitous in life activities. And indispensable. Therefore, accurately identifying the contacts between protein residues has important guiding significance for understanding protein functions, analyzing the relationship between biomolecules, and designing new drugs. [0003] Research literature found that many methods for predicting protein residue contact maps have been proposed, such as: DNCON2 (Adhikari, B., Hou, J. and Cheng, J. (2017) DNCON2: improved protein contact prediction using two-level ...

AI Technical Summary

Problems solved by technology

Although the existing methods can be used to predict the contact map of protein residues, a large number of training data sets and machine learning algorithms are commonly used, so the calculation cost is high, and because the noise information in the training set has not received enough attention, the prediction accuracy is not enough. Guaranteed to be optimal

[0004] To sum up, the existing protein residue contact map prediction methods are still far from the requirements of practical application in terms of calculation cost and prediction accuracy, and urgently need to be improved.

Images