Improved genome-scale multi-target virtual screening via a novel collaborative filtering approach to cold-start problem

Abstract

Conventional one-drug-one-gene approach has been of limited success in modern drug discovery. Polypharmacology, which focuses on searching for multi-targeted drugs to perturb disease-causing networks instead of designing selective ligands to target individual proteins, has emerged as a new drug discovery paradigm. Although many methods for single-target virtual screening have been developed to improve the efficiency of drug discovery, few of these algorithms are designed for polypharmacology. Here, we present a novel theoretical framework and a corresponding algorithm for genome-scale multi-target virtual screening based on the one-class collaborative filtering technique. Our method overcomes the sparseness of the protein-chemical interaction data by means of interaction matrix weighting and dual regularization from both chemicals and proteins. While the statistical foundation behind our method is general enough to encompass genome-wide drug off-target prediction, the program is specifically tailored to find protein targets for new chemicals with little to no available interaction data. We extensively evaluate our method using a number of the most widely accepted gene-specific and cross-gene family benchmarks and demonstrate that our method outperforms other state-of-the-art algorithms for predicting the interaction of new chemicals with multiple proteins. Thus, the proposed algorithm may provide a powerful tool for multi-target drug design.

Predicting serious rare adverse reactions of novel chemicals

“…The matrices of latent ADR and drug preferences ( and , respectively) are found during the standard minimization procedure. For the sake of brevity, we skip technical details (see [20]), but emphasize that the key idea behind our approach is to demand that and are small in one dimension. That way, the output matrix of predicted interaction probabilities (recovered signal) must be of small rank and, in turn, free of noise.…”

“…While we have originally developed and published the analytical framework (1) for the drug-target interaction problem [20], we subsequently noticed that the compressed sensing is much more amenable to predicting adverse drug reactions (ADRs). In contrast to drug-target interaction problem, where the baseline data is already clean but incomplete, the drug-ADR association data is both, incomplete and noisy.…”

Link Prediction in Multi-layer Networks and Its Application to Drug Design

“…They simulate random walk with restart by matrix multiplication, and show that only using a single similarity measure or ignoring the interaction network deteriorates results. Three networks are also used in [7], the authors discuss different options for similarity measures, and perform low-rank matrix factorization on the adjacency/similarity matrices. They address sparsity by giving non-existing links a small non-negative weight.…”

“…A semantically similar problem setting that also faces the sparsity problem is that of product recommendation, and recommender systems have therefore been adapted for the problem setting [7]. A simple recommender system-like approach implements, for instance, the reasoning that if two compounds are both linked with several shared targets, and one of them is linked to an additional one, it is reasonable to assume that the other one should be as well.…”

DDR: efficient computational method to predict drug–target interactions using graph mining and machine learning approaches

“…In our study ( Olayan et al. , 2018 ), we performed 96 computational experiments, including six that are related to the COSINE method ( Lim et al. , 2016 ).…”