Constraint-Based Modeling of Allelic Variation Mechanistically Links Genome-Wide Associations

The advent of genome-sequencing technologies has provided datasets describing the genetic and phenotypic diversity of microbial species at an unprecedented scale. Statistical approaches have led the way in addressing this microbial data deluge, but lack the ability to reason mechanistically about genotype-phenotype associations. However, with the growing availability and scope of genome-scale reconstructions of microorganisms, a mechanistic analysis of these datasets may be possible. Here we realize such an analysis through development and application of an allele-centric constraint-based modeling framework. The framework is applied to a dataset of 1,595 drug-tested M. tuberculosis strains using both reconstruction and inference-based approaches. Known features of M. tuberculosis AMR evolution are recapitulated and new insights revealed.