Chemical genetics in drug discovery

Abstract

Chemical-genetic approaches are based on measuring the cellular outcome of combining genetic and chemical perturbations in large-numbers in tandem. In these approaches the contribution of every gene to the fitness of an organism is measured upon exposure to different chemicals. Current technological advances enable the application of chemical genetics to almost any organism and at an unprecedented throughput. Here we review the underlying concepts behind chemical genetics, present its different vignettes and illustrate how such approaches can propel drug discovery.

Keywords: Drug interactions; Drug resistance; Drug target; Genomics; High-throughput screening.

Figure 1

Basic concepts and approaches in chemical genetics. Chemical-genetic approaches are based on the combination of genetic and chemical perturbations. The fitness of genome-wide libraries of gain-of-function and loss-of-function mutations is assessed upon exposure to large numbers of drugs. Mutant libraries can be pooled or arrayed. In pooled screens barcoded mutants compete among each other after exposure to a certain drug, and their relative abundance is measured by barcode sequencing. In arrayed screens mutants are ordered and their fitness or additional macroscopic phenotypes can be assessed in an independent fashion.

Figure 2

Gene-dosage perturbations reveal drug target. Gene dosage perturbations lead to different insights on a drug MoA depending on the nature of the drug target. In the case of a monomeric protein target, its overexpression determines a right-shift in the growth inhibition curve of the drug (i.e. higher drug concentrations are needed to produce the same growth inhibition). However, if the drug target belongs to a protein complex, a curve shift in comparison with the wildtype one is evident only if the drug can bind to the subunit alone and if the subunit is present/functionally active in the absence of its partners (*). If the protein is a functional target or simple present only as part of the complex, then overexpression does not yield any evident change in the effect of the drug. Overexpression of complex co-members does not change the drug’s inhibition curve. Knockdown perturbations can cause detectable changes in the drug growth inhibition curve (left-shift), both in the case of monomeric and of protein complexes targets. In the case of protein complexes, both the target protein and its complex co-members are shifted.

Figure 3

Cross-resistance and collateral sensitivity maps using publsihed chemical genetics data. Gray edges depict cross-resistance, i.e. gene-drug phenotypic events where mutants are either significantly more susceptible (negative) or more resistant (positive) to both drugs. Red edges depict directional collateral sensitivity- mutants which make cell more resistant to drug A, but more sensitive to drug B, or vice versa. Edge thickness denotes number of mutants. Data used to create network come from published chemical genetics data and drugs selected based on overlap with previous cross-resistance studies , .

Figure 4

Chemical genetics facilitate the mechanistic dissection and prediction of drug-drug interactions. a) Isobologram illustrating different cases of drug interactions. Synergistic, antagonistic and suppressive interactions are represented as phenotypic deviations from the expected additive effect. b) Drug synergies or antagonisms can be profiled in a genome-wide library of mutants. A ε-score assessing the drug-drug interaction is calculated for every mutant, and corresponds to the difference of observed versus expected fitness in the presence of two drugs . Although the vast majority of mutants exhibit a wildtype behavior, in some mutants the drug-drug interaction becomes neutral. These mutants are reflective of the molecular processes the drug interaction depends on. c) Chemical genetics data, drug structural features and previously known drug-drug interaction data can be integrated to predict a more complete drug-drug interaction network. Such networks can be extrapolated to phylogenetically related species.