Predicting drug-target interactions using probabilistic matrix factorization

BalestraWeb: efficient online evaluation of drug-target interactions

SUMMARY

BalestraWeb is an online server that allows users to instantly make predictions about the potential occurrence of interactions between any given drug-target pair, or predict the most likely interaction partners of any drug or target listed in the DrugBank. It also permits users to identify most similar drugs or most similar targets based on their interaction patterns. Outputs help to develop hypotheses about drug repurposing as well as potential side effects.

AVAILABILITY AND IMPLEMENTATION

BalestraWeb is accessible at http://balestra.csb.pitt.edu/. The tool is built using a probabilistic matrix factorization method and DrugBank v3, and the latent variable models are trained using the GraphLab collaborative filtering toolkit. The server is implemented using Python, Flask, NumPy and SciPy.

Adverse drug reaction prediction using scores produced by large-scale drug-protein target docking on high-performance computing machines

Late-stage or post-market identification of adverse drug reactions (ADRs) is a significant public health issue and a source of major economic liability for drug development. Thus, reliable in silico screening of drug candidates for possible ADRs would be advantageous. In this work, we introduce a computational approach that predicts ADRs by combining the results of molecular docking and leverages known ADR information from DrugBank and SIDER. We employed a recently parallelized version of AutoDock Vina (VinaLC) to dock 906 small molecule drugs to a virtual panel of 409 DrugBank protein targets. L1-regularized logistic regression models were trained on the resulting docking scores of a 560 compound subset from the initial 906 compounds to predict 85 side effects, grouped into 10 ADR phenotype groups. Only 21% (87 out of 409) of the drug-protein binding features involve known targets of the drug subset, providing a significant probe of off-target effects. As a control, associations of this drug subset with the 555 annotated targets of these compounds, as reported in DrugBank, were used as features to train a separate group of models. The Vina off-target models and the DrugBank on-target models yielded comparable median area-under-the-receiver-operating-characteristic-curves (AUCs) during 10-fold cross-validation (0.60-0.69 and 0.61-0.74, respectively). Evidence was found in the PubMed literature to support several putative ADR-protein associations identified by our analysis. Among them, several associations between neoplasm-related ADRs and known tumor suppressor and tumor invasiveness marker proteins were found. A dual role for interstitial collagenase in both neoplasms and aneurysm formation was also identified. These associations all involve off-target proteins and could not have been found using available drug/on-target interaction data. This study illustrates a path forward to comprehensive ADR virtual screening that can potentially scale with increasing number of CPUs to tens of thousands of protein targets and millions of potential drug candidates.

Drug-disease association and drug-repositioning predictions in complex diseases using causal inference-probabilistic matrix factorization

The high incidence of complex diseases has become a worldwide threat to human health. Multiple targets and pathways are perturbed during the pathological process of complex diseases. Systematic investigation of complex relationship between drugs and diseases is necessary for new association discovery and drug repurposing. For this purpose, three causal networks were constructed herein for cardiovascular diseases, diabetes mellitus, and neoplasms, respectively. A causal inference-probabilistic matrix factorization (CI-PMF) approach was proposed to predict and classify drug-disease associations, and further used for drug-repositioning predictions. First, multilevel systematic relations between drugs and diseases were integrated from heterogeneous databases to construct causal networks connecting drug-target-pathway-gene-disease. Then, the association scores between drugs and diseases were assessed by evaluating a drug's effects on multiple targets and pathways. Furthermore, PMF models were learned based on known interactions, and associations were then classified into three types by trained models. Finally, therapeutic associations were predicted based upon the ranking of association scores and predicted association types. In terms of drug-disease association prediction, modified causal inference included in CI-PMF outperformed existing causal inference with a higher AUC (area under receiver operating characteristic curve) score and greater precision. Moreover, CI-PMF performed better than single modified causal inference in predicting therapeutic drug-disease associations. In the top 30% of predicted associations, 58.6% (136/232), 50.8% (31/61), and 39.8% (140/352) hit known therapeutic associations, while precisions obtained by the latter were only 10.2% (231/2264), 8.8% (36/411), and 9.7% (189/1948). Clinical verifications were further conducted for the top 100 newly predicted therapeutic associations. As a result, 21, 12, and 32 associations have been studied and many treatment effects of drugs on diseases were investigated for cardiovascular diseases, diabetes mellitus, and neoplasms, respectively. Related chains in causal networks were extracted for these 65 clinical-verified associations, and we further illustrated the therapeutic role of etodolac in breast cancer by inferred chains. Overall, CI-PMF is a useful approach for associating drugs with complex diseases and provides potential values for drug repositioning.

Drug-target interaction prediction via chemogenomic space: learning-based methods

INTRODUCTION

Identification of the interaction between drugs and target proteins is a crucial task in genomic drug discovery. The in silico prediction is an appropriate alternative for the laborious and costly experimental process of drug-target interaction prediction. Developing a variety of computational methods opens a new direction in analyzing and detecting new drug-target pairs.

AREAS COVERED

In this review, we will focus on chemogenomic methods which have established a learning framework for predicting drug-target interactions. Learning-based methods are classified into supervised and semi-supervised, and the supervised learning methods are studied as two separate parts including similarity-based methods and feature-based methods.

EXPERT OPINION

In spite of many improvements for pharmacology applications by learning-based methods, there are many over simplification settings in construction of predictive models that may lead to over-optimistic results on drug-target interaction prediction.