Fig. 2

YARS protein expression is validated as a predictor of chemotherapeutic response in breast cancer. a Comparison of the predictive powers of the combinations of the best candidate proteins determined using machine learning algorithms for predicting chemotherapy response. The predictive performance of each algorithm to predict chemotherapy response is represented by a linear plot along with AUC, sensitivity, and specificity (red square, upregulated proteins in the nCR group; blue square, upregulated proteins in the CR group). b A convergent filtering approach utilizing predictive values to select the candidate with potential therapeutic value. c Needle biopsy FFPE tissues from a set of 123 independently enrolled patients were used to verify the predictive performance of KIAA1522 or YARS; 79 patients were enrolled to confirm the predictive performance of PDCD6 for all intrinsic subtypes of breast cancer. Subset analysis based on intrinsic subtypes, including luminal, HER2-enriched, and triple-negative breast cancer (left). The heatmap demonstrates negative to gradient staining, representing the H-score values for proteins. The graph provided next to the heatmap demonstrates an intensity ratio of immunohistochemical staining between the nCR and CR groups (right, Mann-Whitney U test). CR, complete remission; nCR, non-complete remission