Method for predicting drug targeting protein-to-protein interaction based on multivariate data

Abstract

The invention discloses a method for predicting drug targeting protein-to-protein interaction (PPI) based on multivariate data. According to the method, with the judgment whether a drug is targeted to protein-to-protein interaction as the research object, an integrated classifier algorithm for predicting the drug targeting protein-to-protein interaction based on eleven characteristics such as the structure, function and amino acid sequence of protein is developed. The prediction method can be used for predicting whether interaction exists between protein and protein or not and whether the interaction target of a drug and protein occurs in the protein binding interface (PPI interface) or not.

Description

technical field [0001] The invention relates to a prediction method for predicting the interaction of drug-targeted proteins under the background of machine learning and bioinformatics knowledge, in particular to a prediction method for predicting the interaction of drug-targeted proteins based on multivariate data. Background technique [0002] Protein interaction (Protein Protein Interaction, PPI) plays a central role in life activities. Studies have proved that protein interaction is drug-targeted. The study of protein interaction provides an important theoretical basis for drug target discovery and drug design. For example, once the amount of the apoptosis inhibitor XIAP in the cells exceeds the normal value, the cells will not undergo normal apoptosis, which in turn will cause tumors and cancers. However, XIAP can inhibit apoptosis by interacting with caspase-9, an activating intracellular caspase. Therefore, inhibiting the interaction between XIAP and caspase can achie...

AI Technical Summary

Problems solved by technology

[0007] The above-mentioned methods have the following disadvantages. First, there are few types of data, that is, only one or two types of data are considered for fitting, and the prediction effect is not accurate enough.

Second, the number of features used is limited and the prediction accuracy is not enough

The second is that in the calculation process, the prediction method combined with machine learning does not adopt effective feature selection, resulting in samples containing a large amount of redundant information, which cannot achieve accurate prediction results

Images