Systematic prediction of drug combinations based on clinical side-effects

Abstract

Drug co-prescription (or drug combination) is a therapeutic strategy widely used as it may improve efficacy and reduce side-effect (SE). Since it is impractical to screen all possible drug combinations for every indication, computational methods have been developed to predict new combinations. In this study, we describe a novel approach that utilizes clinical SEs from post-marketing surveillance and the drug label to predict 1,508 novel drug-drug combinations. It outperforms other prediction methods, achieving an AUC of 0.92 compared to an AUC of 0.69 in a previous method, on a much larger drug combination set (245 drug combinations in our dataset compared to 75 in previous work.). We further found from the feature selection that three FDA black-box warned serious SEs, namely pneumonia, haemorrhage rectum, and retinal bleeding, contributed mostly to the predictions and a model only using these three SEs can achieve an average area under curve (AUC) at 0.80 and accuracy at 0.91, potentially with its simplicity being recognized as a practical rule-of-three in drug co-prescription or making fixed-dose drug combination. We also demonstrate this performance is less likely to be influenced by confounding factors such as biased disease indications or chemical structures.

Using Big Data to Discover Diagnostics and Therapeutics for Gastrointestinal and Liver Diseases

“…There are several potential advantages of drug combinations: better coverage of multiple disease mechanisms than with a single agent; 108 dose reduction for a potentially toxic component of the combination while maintaining therapeutic efficacy; 109 synergistic effects, including synthetic lethality in cancer 110 ; and the prevention of innate and acquired drug resistance. 111 Efficient, high-throughput identification of effective drug combinations is therefore an important component of a successful therapeutic discovery pipeline.…”

“…111 Efficient, high-throughput identification of effective drug combinations is therefore an important component of a successful therapeutic discovery pipeline. To this end, a number of methods for computational prediction of synergistic compounds have been developed, 112 based on side-effect profiles, 108 chemical and pathway data, 113 network analyses, 114 and drug-induced gene expression patterns. 109 Curated databases of reported drug combinations and other resources are also publicly available.…”

Diffusion mapping of drug targets on disease signaling network elements reveals drug combination strategies

“…sequence, chemical structure), and to evaluate the approach in a pan-cancer setting. Another interesting aspect to explore in future work would be the impact of drug community structure and the similarity of toxicity profiles to derive drug combination models that balance efficacy and toxicity simultaneously [19]. Finally, it will be important to confirm predicted mechanisms of drug synergy in prospective in vitro experiments (e.g., CRISPR gene editing) to assess the impact of local paths and gene sub-networks in the overall disease signaling network.…”

Computational chemical biology and drug design: Facilitating protein structure, function, and modulation studies

“…Many statistical learning algorithms have been used in drug‐combination discovery. For example, Huang et al . have collected clinical SEs from postmarketing surveillance and drug labels and have built different ML models to identify the three main features contributing to unsafe drug combinations, that is, pneumonia, rectal hemorrhage, and retinal bleeding.…”