Fig. 4

Detection of colon adenocarcinoma with a 6-gene marker panel. To identify a minimal marker panel, cumulative DNA methylation data using QM-MSP from fresh frozen tissues of colon adenocarcinoma and adjacent normal tissues were re-analyzed. A panel of 6 genes, a subset of the 13-gene panel shown in Fig. 3, was selected using criteria described in Table 2 legend. The percent methylation (%M) of each gene, and the cumulative methylation index (CMI; the sum of %M for all genes in the panel) were determined for each sample. Box–whiskers plot: In training (A) and test samples (D) significantly higher methylation was observed in samples of carcinoma compared to adjacent normal tissues (P < 0.0001; Mann–Whitney). Histogram plot: Data from samples in the training set (B) and Test set (E) are plotted. For each sample (X-axis), the height of the histogram bar indicates the level of cumulative methylation (Y-axis), each colored segment represents an individual gene, and the size of the segment is proportional to the %M of that gene. Microsatellite instability (MSI) status for each carcinoma is indicated by the bar map below the X-axis: blue (microsatellite unstable), black (microsatellite stable), and gray (unknown). Receiver operating characteristic (ROC): The ROC curve from the training set (C), carcinoma versus normal control, identified a laboratory threshold maximized for sensitivity for detecting carcinoma at a specificity > 90%. This threshold (CMI = 26.0) was locked and used to determine the assay sensitivity and specificity for the test set (F)