Fig. 5: Validation using independent datasets.
From: Deep generative neural network for accurate drug response imputation
A Survival analysis of samples from GSE33072 who were treated with erlotinib. The P value was from a log-rank test comparing two groups of samples defined by the predicted response to erlotinib (HR: high response, greater than the median; LR: low response). B Comparison of predicted response to erlotinib in cell lines stratified into epithelial-like (n = 44) or mesenchymal-like (n = 25) group using the dataset GSE32989. The P value was obtained by using a two-sided t test. C, E Survival analysis of melanoma samples treated with vemurafenib (data from GSE65185). These samples were carriers with BRAF V600 mutations. The P value was from a log-rank test comparing two groups of samples defined by the predicted response to PLX4720 (HR: high response, greater than the median; LR: low response). D, F Comparison of predicted response to PLX4720 in parental cell lines and in derived resistant cell lines. G Comparison of predicted response to paclitaxel between BRCA subgroups with pCR (pathological complete response, npCR = 122 in GSE25055, npCR = 27 in GSE32646, and npCR = 56 in GSE20194) and RD (residual disease, also called nCR or non-pCR in GSE32646 (nRD = 188 in GSE25055, nnCR = 88 in GSE32646, nRD = 222 in GSE20194). In all panels, the P value was obtained by using a two-sided t test. In the boxplots (B, G), each box shows the interquartile range (IQR between Q1 and Q3) for the corresponding set. The central mark (horizontal line) shows the median and the dots show the rest of the distribution based on IQR [Q1−1.5 × IQR, Q3 + 1.5 × IQR].
