Fig. 4: Mitochondrial phenotype composition of individual features and principal components.
a. Twenty of the most discriminative features between MFN2V459fs/WT positive controls and No Cas9 (No Dox) negative controls in the mitochondrial gene knockout screening data as determined by the Kolmogorov-Smirnov test statistic. Bars are color-coded according to the wavelength from which each feature was derived. Every bar shown had a KS p-value near 0. b–e Cumulative probability histograms from the full dataset with the red line showing MFN2V459fs/WT cells and the blue line showing No Cas9 cells. The stated AUCs are for the full dataset (training and test). b Cumulative probability histogram segmented according to the most discriminative feature: MitoTracker nuclei total intensity. c The second most discriminative feature: MitoTracker cell intensity spreading. d. The product of the top two features. e One of the 12-feature logistic regressions from the Mito Path ScoreR ensemble. f, g Scatter plots and the Pearson’s R correlation between the Mito Path ScoreR and each of the top three principal components for MitoTracker and TMRM. The top loadings for each principal component are included. The percentage listed after the PC1, 2, and 3 are the percent explained variance for that PC.
