Extended Data Fig. 3: Feature weights of Lasso classifiers. | Nature

Extended Data Fig. 3: Feature weights of Lasso classifiers.

From: Large-scale chemical–genetics yields new M. tuberculosis inhibitor classes

Extended Data Fig. 3

a, Feature weights of the hypomorphs used by the Lasso binary classifier to predict DNA gyrase inhibitors. We trained on the fluoroquinolones, with the GyrA hypomorph being the most predictive strain. Features are denoted as the target gene of the hypomorph and compound concentration separated by an underscore. b, As in a for the Lasso binary classifier to predict inhibitors of mycolic acid biosynthesis. We trained on known InhA inhibitors, with the MshC hypomorph being a prominent discriminator. c, As in a for the Lasso binary classifier to predict inhibitors of folate biosynthesis. We trained on the sulfonamides, with the TrpG hypomorph being a prominent discriminator. d, As in a but trained on chemical–genetic interaction profiles from the sulfonamides and confirmed new folate biosynthesis inhibitors. e, As in a but trained on chemical–genetic interaction profiles of the newly identified inhibitors of tryptophan biosynthesis.

Back to article page