Drug molecule recommendation system for regulating and controlling disease targets based on a deep learning, computer equipment and storage medium

Abstract

The invention discloses a drug molecule recommendation system for regulating and controlling disease targets based on a deep learning method, and belongs to the technical field of drug relocation, convolutional neural networks and residual networks. The system comprises a deep residual error network model, and the deep residual error network model comprises an embedded network, a convolution residual error neural network and a full connection layer residual error network. The embedded network converts a drug molecule SMILES sequence or a protein amino acid sequence into a binary matrix. The convolution residual neural network comprises three convolutional layers, an addition layer and a maximum pooling layer, and is a network represented by a 'learning' drug molecule SMILES sequence or protein amino acid sequence feature, the input of the network is a binary matrix representing a drug molecule or protein, and the output of the network is a feature representation vector of the drug molecule or protein. The full connection layer residual network comprises three full-connection layers, two dropout layers and an addition layer, the input is a splicing vector expressed by drug moleculeand protein characteristics and an actual binding affinity value of the drug molecule and the protein characteristics, and the output is a binding affinity prediction value of the drug molecule and the protein.

Description

technical field [0001] The invention relates to the technical field of deep learning and drug repositioning, in particular to a drug molecule recommendation system, computer equipment, and storage media based on deep learning methods for regulating disease targets. Background technique [0002] Since the outbreak of the new crown epidemic, due to the long cycle of traditional drug development and high failure rate, drug repositioning (that is, new use of old drugs) has become a common drug development strategy in the face of public health emergencies, and it is also the main way to solve the plight of new drug development. One of the methods. Diseases are related to multiple targets, and most of the current drugs are developed for a single target. Therefore, finding a drug that regulates a single target can achieve the purpose of treating diseases related to this target disease. Traditional Chinese medicine is a complex mixture. Each Chinese medicine contains multiple compo...

AI Technical Summary

Problems solved by technology

As the model continues to deepen, the information is continuously compressed; on the other hand, with the use of the dropout layer, although it can effectively prevent the model from overfitting, it also leads to different recommendation results each time, and the most effective drug molecule , and not necessarily the one with the best predicted value in the model with the highest accuracy

Images