Fig. 5: Characterization of isoniazid MACs.
From: A biochemically-interpretable machine learning classifier for microbial GWAS
a Horizontal bar plots of pathways enriched with significant isoniazid-associated fluxes with FDR < 0.05. b Boxplots of katG, nuoL, ndh, gpdA2, accD6, fabD, kasA, and accD5 allele-specific fluxes for the reactions catalyzed by their gene-products. Alleles are rank ordered from least to greatest by their log odds ratio (LOR), from left to right. The flux (y-axis) is the median scaled flux across the high-quality isoniazid MACs. The boxes are colored according to the allele LOR, where positive corresponds to resistant (R) dominant while negative corresponds to susceptible (S) dominant. The box shows the quartiles of the dataset with the median noted as the horizontal line while the whiskers extend to show the rest of the points with outliers colored gray. The R-squared and P-value for the regression between allele LOR and flux is noted. See Supplementary Data 5 for list of mutations per allele. c Clustered heatmap of allele LOR-flux correlations for significant reactions in TCA Cycle V, Oxidative phosphorylation, and Mycolate biosynthesis. d Pathway depiction of TCA Cycle V, Oxidative phosphorylation, and Mycolate biosynthesis. Significant allelic effects are shown by dashed lines and colored for gpdA2 and nuoL.
