Fig. 4

The confounding clinical variables to the iGenSig-Rx prediction model. A, B The interactions of confounding clinical variables were assessed based on multiple logistic regression models for therapeutic response predictive values in ACOSOG Z1041 and NSABP B-41. The bar plot shows the -log10 transformed p values of Chi-Square tests comparing the pair-wise multiple logistic regression models with the simple logistic regression model. C The predictive values of the iGenSig-Rx model are displayed for ACOSOG Z1041 and B-41 subjects stratified by tumor stages; stage I-III, stage I-II or stage III. The predictive values were measured in pCR or non-pCR subjects separately in each tumor stage stratification. CALGB 40601 stage II-III (n = 277) or CALGB 40601 stage III (n = 89) were used as the train sets to build the independent iGenSig-Rx modeling. ACOSOG Z1041 and NSABP B-41 are used as the test sets