FIGURE 2.

Prediction of pCR patients using PPCET. Receiver operating characteristic (ROC) curves for the training cohort (A), validation cohort (C), and all cohorts (E). Performance of classifiers for the training cohort (B), validation cohort (D), and all cohorts (F). In total, 194 LARC patients with neoadjuvant therapy were used to develop classifiers to predict pCR patients. The classifier PPCET with the largest area under curve (AUC) was identified (AUC = 0.89 [0.83‐0.94]). Machine learning analysis of the variables of the global nucleosome, including fragment profiles of 5 Mb windows (5Mb), subcompartments of genome (Sub), and mitochondrial DNA copy number (mtDNA), was implemented