Figure 2

Classification analyses show early neoplasia is distinct from normal tissue and cancer. (A) Cross-validated probabilities from the unpaired three-class PAM analysis predicting the diagnostic class of each sample. Classes are normal, early neoplasia, and cancer (includes DCIS and IDC). Samples are arranged as in Table S3 in Additional file 1. The pathologist-assigned diagnosis is listed above. The circles represent the predictions based on the gene expression of the 44 selected genes. The probabilities for each sample sum to 1. (B) Mean-centered, normalized heatmap showing 3SEQ transcript levels for each of the 44 genes selected in the three-class PAM analysis in (A). Samples are arranged as in (A). Red is high expression; green is low expression. (C) Cross-validated probabilities from the unpaired two-class (normal versus early neoplasia) PAM analysis for predicting the diagnostic class of each sample. Normal and early neoplasia samples are arranged as in Table S3 in Additional file 1. The pathologist-assigned diagnosis is listed above. The circles represent the predictions based on the gene expression of the 180 selected genes. The probabilities for each sample sum to 1. (D) Mean-centered, normalized heatmap showing 3SEQ transcript levels for each of the 180 genes selected in the two-class PAM analysis in (C). Samples are arranged as in (C). Red is high expression; green is low expression.