Abstract:

A new prediction method for proteinprotein interaction (PPI) was proposed based on an improved pseudo amino acid composition (PseAA) feature model and random forest. A new PseAA feature model based on the Geary autocorrelation function is used to evaluate amino acid properties related to PPI. Then accordingto the results of evaluation, relevant properties are selected to integrate together by another new PseAA feature model based on the Minkowski function. The randomforest is adopted as classifier for learning and prediction. The results obtained in the experiment indicate that this method can improve accuracy.

Key words: proteinprotein interaction; pseudo amino acid composition;random forest