Fig. 4: Drug-specific predictions of PRIME. | npj Systems Biology and Applications

Fig. 4: Drug-specific predictions of PRIME.

From: Quantitative prediction of conditional vulnerabilities in regulatory and metabolic networks using PRIME

Fig. 4

a Heatmap of PRIME derived fitness for all TF knockouts in the presence of seven primary drugs and control at 24 h. The numbers indicate the concentration of drug used in μg/mL. INH isoniazid, BDQ bedaquiline, RIF rifampicin, LZD linezolid, MOX moxifloxacin, CFZ clofazamine, PA824 pretomanid. b Sensitivity and specificity of PRIME, PROM, and IDREAM predicted TF essentiality in the presence of INH as determined by LOOCV analysis for the area under the receiver operating characteristic curve (ROC AUC). Statistical significance was calculated as p-value with two-sample t-test. ****: p-value < 0.0001. All the boxes in the boxplot indicate the upper and lower quartiles of the data and the middle line is the median with the whiskers extending to 1.5× interquartile range. c Correlation of TnSeq experimental fitness ranking of TFs and PRIME derived fitness ranks. d BioTapestry visualization showing a subset of the gene regulatory network of Mtb with PRIME predictions during INH treatment. Some of the highlighted TFs were predicted as essential in the presence of INH (Rv0827c, Rv1049, and Rv0472c), while others were predicted essential in both the absence and presence of INH (Rv1423, Rv1828, Rv3246c, and Rv2610c). The lightened TFs were predicted essential in the untreated control but non-essential in the presence of INH (Rv3681c, Rv1816, and Rv0576). All of these PRIME predictions were validated by experimental fitness screening in relevant conditions.

Back to article page