METHOD AND DEVICE FOR PREDICTING INTERACTION RELATIONSHIP BETWEEN MEMBRANE PROTEIN RESIDUES

摘要

Disclosed is a method for predicting the interaction relationship between membrane protein residues, comprising: obtaining a membrane protein, where the protein structure of same has been analyzed to be a training set; extracting the features of the unbalanced classification for distinguishing interacting residue pairs from non-interacting residue pairs in the membrane protein where the protein structure has been resolved; training a predictive model with the extracted features of the unbalanced classification by means of a SMOTE-Boost algorithm; and predicting the interaction relationship between membrane protein residues where the protein structures are unknown according to the trained predictive model. Due to the use of the features of the unbalanced classification to train the predictive model, the trained predictive model can avoid the loss of useful information, and can help to improve the accuracy and coverage of the prediction.