Fig. 4: Performance and accuracy of DRUML to rank drugs based on efficacy.
From: Drug ranking using machine learning systematically predicts the efficacy of anti-cancer drugs
a Total number of models generated from phosphoproteomics, proteomics and RNA-seq data. Input data were split into solid (n = 22) and AML (n = 26) cell line groups for model building. b Validation errors for each model generated binned by ML method and input dataset. c Comparison between measured and predicted drug response values produced by DL analysis of phosphoproteomics distance values in the AML datasets. Each data-point represents a drug prediction color coded by the drugs’ mode of action. Each cell line was analyzed in triplicate and representative comparisons are shown. Dotted line denotes slope of 1 with 0 intercept. d Spearman rank correlation coefficient rho values between predicted and actual responses returned by the different learning models and input datasets). Boxplots (b and d) have median centers, upper and lower quartile box boundaries and range upper and lower hinges. Learning algorithms were random forest (rf), cubist, bayesian estimation of generalized linear models (bglm), partial least squares (pls), principal component regression (pcr), support vector machine (svm), deep learning (dl) and neural network (nnet).
