Fig. 2: Evaluation of the discriminative models for mutant cells.
From: Machine learning approach for discrimination of genotypes based on bright-field cellular images
a Receiver operating characteristic (ROC) curves in the discrimination of PSMB5 mutant cells from wild-type cells. Thin blue lines represent ROC curves calculated with tenfold cross-validation. Thick blue line and gray area represent the mean and the quartile of the ROC curves, respectively. b Relationship between posterior probabilities and linear predictors of the discriminative model for PSMB5 mutant cells. Upper and middle plots show the distributions of linear predictors of wild-type cells and mutant cells, respectively. Lower plot shows a sigmoid curve between posterior probabilities and linear predictors. c AUC of the discriminative model for each mutant cell type. The error bar shows the standard deviation in tenfold cross-validation. Red dashed line represents AUC = 0.59.
