A Random Projection Ensemble Approach to Drug-Target Interaction Prediction

Abstract

AbstractDrug-target interaction prediction is very important in drug development. Since determining drug-target interactions is costly and time-consuming by experiments, it is a complement to determine the interactions by computational method. To address the issue, a random projection ensemble approach is proposed and drug-compounds are encoded with feature descriptors by software “PaDEL-Descriptor”, while target proteins are encoded with physicochemical properties of amino acids. From 544 properties in AAindex1, 34 relatively independent physicochemical properties are extracted. Random projection on the vector of drug-target pair with different dimensions can map the original space onto a reduced one and thus yield a transformed vector with fixed dimension. Several random projections build an ensemble REPTree system. Experimental results showed that our method significantly outperformed and ran faster than other state-of-the-art drug-target predictors.KeywordsRandom projectionDrug-target interactionREPTreeEnsemble system