Extended Data Fig. 2: Quality control of the main interaction screen and assessment of fitness calculation methods. | Nature Microbiology

Extended Data Fig. 2: Quality control of the main interaction screen and assessment of fitness calculation methods.

From: Systematic analysis of drug combinations against Gram-positive bacteria

Extended Data Fig. 2

a-b, Donor (a) and recipient (b) drug fitness correlation between biological replicates. Pearson correlation (R = 0.84 and 0.89, respectively, p < 2.2e-16) is calculated between biological replicates, corresponding to different experimental runs/batches. P-values are obtained from a two-tailed one-sample t-test assessing the significance of the Pearson correlation (\({H}_{0}:\{t=0,{R}=0\)}). c, Technical replicate correlation. Pearson correlation is calculated between replicate wells within the same plate for combination plates (where donor drugs were added) and control recipient plates (where no donor drug was added), for the four strains screened. Plates for which technical plate correlation was < 0.7 (red) were removed from the data. Box limits correspond to first and third quartiles, with the median marked, and whiskers to the most extreme data points up to 1.5 times the IQR. d, Performance of endpoint OD- and AUC-based measurements against the benchmarking set. Precision-recall curves are shown for q-value intervals increasing by 0.01. Curves highlighted correspond to the effect-size cut-off selected for the screen (interaction score = |0.1|). The significance cut-off (FDR < 0.05) is marked. e-g, Endpoint-OD- and AUC-based fitness values for all strains are highly correlated (e; Pearson correlation, R = 0.96, p < 2.2e-16, obtained from a two-tailed one-sample t-test assessing the significance of the Pearson correlation, n = 270189), whereas fitness values only based on growth rate correlated worse with either AUC- (f) or OD-based (g) fitness values (Pearson correlation, R = 0.68 and 0.75, respectively, p < 2.2e-16, obtained as in e, n = 270189). h-i, Comparison between estimated and experimentally measured single-drug fitness for donor (h) (Pearson correlation, R = 0.98, p < 2.2e-16, obtained as in e, n = 5208) and recipient (i) drugs (Pearson correlation, R = 0.96, p < 2.2e-16, obtained as in e, n = 1718).

Source data

Back to article page