Multi-attention method for predicting drug target interactions

Abstract

The invention discloses a method for predicting drug target interaction, and the method is used for predicting the affinity score of a drug target pair based on an end-to-end model (MATT-DTI) of a multi-attention module. The model is wanted by two self-attention module distribution modeling drugs and interaction of drug target pairs. MATT-DTI, as shown in Figure 1, firstly, features of a drug sequence are extracted by using a related self-attention model and a convolutional neural network through a correlation relationship between coding atoms, so that deep expression of a drug is obtained; secondly, MATT-DTI obtains deep expression of a target sequence through coding of a convolutional neural network model; and finally, establishing a module based on a multi-head self-attention mechanism, and extracting interaction information of a drug target pair through similarity modeling between drug deep expression and target deep expression so as to obtain a predicted affinity score.

Description

technical field [0001] The invention relates to the technical field of computer multimedia, in particular to a multi-attention method for predicting drug-target interaction. Background technique [0002] In real-world scenarios, the task of drug-target interaction prediction has received increasing attention due to its important role in the drug discovery process. Methods for drug-target interaction prediction can be considered as classification tasks and regression tasks. The classification task is to predict whether there is an interaction between drug-target pairs. The regression task is to predict the degree of interaction between drug-target pairs, usually measured by affinity scores. Most drug-target interaction prediction methods ignore the atom-to-atom correlations of compounds. For example, DeepDTA, which encodes drugs and targets through a character-level dictionary, not only encodes multi-character atoms into multiple characters, but also ignores the correlatio...

AI Technical Summary

Problems solved by technology

For example, DeepDTA, which encodes drugs and targets through character-level dictionaries, not only encodes multi-character atoms into multiple characters, but also ignores the correlation between atoms and atoms, which leads to the fact that it may not capture nonlinear high-level semantic information

Images