Predicting Drug-Target Interactions for New Drug Compounds Using a Weighted Nearest Neighbor Profile

Abstract

In silico discovery of interactions between drug compounds and target proteins is of core importance for improving the efficiency of the laborious and costly experimental determination of drug-target interaction. Drug-target interaction data are available for many classes of pharmaceutically useful target proteins including enzymes, ion channels, GPCRs and nuclear receptors. However, current drug-target interaction databases contain a small number of drug-target pairs which are experimentally validated interactions. In particular, for some drug compounds (or targets) there is no available interaction. This motivates the need for developing methods that predict interacting pairs with high accuracy also for these 'new' drug compounds (or targets). We show that a simple weighted nearest neighbor procedure is highly effective for this task. We integrate this procedure into a recent machine learning method for drug-target interaction we developed in previous work. Results of experiments indicate that the resulting method predicts true interactions with high accuracy also for new drug compounds and achieves results comparable or better than those of recent state-of-the-art algorithms. Software is publicly available at http://cs.ru.nl/~tvanlaarhoven/drugtarget2013/.

Drug-target interaction prediction: A Bayesian ranking approach

“…The proposed BRDTI method was compared with four state of the art approaches: BLM-NII [25], WNN-GIP [28], NetLapRLS [26] and CMF [30]. Grid-search was used to tune methods' hyperparameters, details can be found in supplementary materials.…”

“…Machine learning methods in general leverage features based on the structure of drugs and targets (e.g., [9,[19][20][21]), drugs' side-effects [22], and the knowledge of already confirmed DTIs [23][24][25][26][27][28][29][30][31][32]. In particular, in case of Bipartite Local Models (BLM) [23] and its extensions (e.g.…”

“…Van Laarhoven et al [27] proposed to use regularized least squares with Gaussian interaction profile kernel (GIP). The method was later improved by incorporating weighted nearest neighbors to be able to predict interactions also for new drugs and targets (WNN-GIP) [28].…”

Fast Kronecker Product Kernel Methods via Generalized Vec Trick

“…• K-nearest neighbors (KNN): KNN methods have enjoyed substantial popularity in graph prediction applications such as biological interaction prediction [63] and recommender systems [64]. The method can model highly nonlinear functions and can scale well especially to low-dimensional problems by using efficient data structures for speeding up the neighborhood-search.…”

A hybrid approach for hot spot prediction and deep representation of hematological protein – drug interactions

“…The AC descriptors used for the interaction between residues of a protein sequence and the ELM at a distance is an accurate and rapid method of learning. Laarhoven [8] was recommended that WNN-GIP be administered in combination with a new compound or target. Computation is done through a regularized least-squares algorithm that combines core products.…”