Prediction of Synergism from Chemical-Genetic Interactions by Machine Learning

Abstract

Summary The structure of genetic interaction networks predicts that, analogous to synthetic lethal interactions between non-essential genes, combinations of compounds with latent activities may exhibit potent synergism. To test this hypothesis, we generated a chemical-genetic matrix of 195 diverse yeast deletion strains treated with 4915 compounds. This approach uncovered 1221 genotype-specific inhibitors, which we termed cryptagens. Synergism between 8128 structurally disparate cryptagen pairs was assessed experimentally and used to benchmark predictive algorithms. A model based on the chemical-genetic matrix and the global genetic interaction network failed to accurately predict synergism. However, a combined random forest and Naïve Bayesian learner that associated chemical structural features with genotype-specific growth inhibition had strong predictive power. This approach identified previously unknown compound combinations that exhibited species-selective toxicity towards human fungal pathogens. This work demonstrates that machine learning methods trained on unbiased chemical-genetic interaction data may be widely applicable for the discovery of synergistic combinations in different species.

Antibiotic efficacy — context matters

“…Mass spectrometry allows interrogation into spatial determinants of antibiotic efficacy [110,111] and synthetic biology has contributed novel tools for perturbing essential genes [112]. Current quantitative models are providing mechanistic insights into several nonlinear components of antibiotic lethality [56,113] and are useful for predicting multidrug treatment outcomes [114,115] and antibiotic drug synergies [116,117]. Integration of these tools will enable identification of additional mechanistic details between primary target inhibition and subsequent lethality.…”

Combinatorial Drug Screening Identifies Ewing Sarcoma–specific Sensitivities

“…Compensatory pathways can remodel the signaling landscape and cause drug ineffectiveness (53). However, synergistic activity of two drugs with distinct primary mechanism of action, can improve treatment efficacy, especially in complex diseases where the control is more likely to be accomplished by using multiple interventions (54,55). By charting disease specific functional synergies onto pathways, we aimed to decipher synergistic signaling cross-talks in a particular Ewing sarcoma context.…”

Systematic chemical-genetic and chemical-chemical interaction datasets for prediction of compound synergism

“…By definition, the biological activity of such compounds would not be detected in many high-throughput screens used in modern drug discovery. Compounds with such latent activities have been termed cryptagens or dark chemical matter 10,11 .…”

“…We recently generated a systematic chemical-genetic dataset in S. cerevisiae to allow the discovery and prediction of synergistic interactions between cryptagens that do not have obvious effects on cell proliferation on their own 11 . Various algorithmic approaches have been developed to predict synergistic compound combinations 1,12,13 .…”

“…The dearth of fully factorial drug combination data matrices has hampered the systematic testing and comparisons of different predictive approaches 1 . To address this shortfall, we generated two large-scale data sets: a chemical-genetic matrix (CGM) of 356,500 pairwise chemical-gene interaction tests and a derived cryptagen matrix (CM) of 8,128 chemical-chemical interaction tests 11 . Based on this data, we developed a machine learning approach that integrates structural features of compounds with chemical-genetic interactions to predict compound synergism 11,15 .…”