Association prediction method for drugs and pathways of knowledge graph attention network

Abstract

The invention discloses a drug and pathway association prediction method for a knowledge graph attention network, and the method comprises the steps: S1, building a drug-pathway knowledge graph of drugs, pathways, genes and diseases according to the data of a database; s2, aggregating multi-hop neighbor information by using a graph attention neural network, and obtaining embedding; s3, inputting the embedding into a multi-layer perceptron, performing prediction scoring, and optimizing model parameters according to a loss function; the data is from a KEGG database and a CTD database. According to the method, a brand new perspective is provided for problem research on drug and pathway association prediction, the knowledge graph and graph attention mechanism technology is applied to drug and pathway association prediction for the first time, a high-order relation is brought into a prediction model, and all related parameters are customized to optimize a prediction target; and a detailed knowledge graph of drug-pathway-gene-disease is constructed, and under a graph neural network framework, modeling of a high-order relationship is realized in an explicit and end-to-end manner, so that the accuracy of the model is improved.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a method for predicting the association between drugs and pathways in a knowledge graph attention network. Background technique [0002] Drug development is a long, complex and difficult process. According to estimates, it costs hundreds of millions of dollars to develop a new drug. Traditional drug development follows the "one drug, one target" approach, which has problems of high research cost and time investment, ignoring the complex mechanism of drug action and the systemic nature of human diseases. Many diseases occur due to multiple gene associations and complex biological pathways, rather than a single gene product. Furthermore, the actions of drugs can affect the complex connections between many related biological pathways. Pathway-based strategies, which can provide information on upstream or downstream genes associated with the target, are widely used to pred...

AI Technical Summary

Problems solved by technology

However, there are still certain accuracy problems in this method based on feature fusion and using artificially constructed features to express drug information and pathway information.

Images