Unsupervised analyses and classification results of candidate CEBPA^dm cases with their GEP and their molecular characteristics. (A) Pairwise correlations between the 674 AML cases (supplemental Table 1). The cells in the visualization are colored by Pearson correlations values, depicting higher positive (red) or negative (blue) correlations, as indicated by the scale bar. CEBPA^sm, CEBPA^dm, CEBPA^{C-terminal mutation}, CEBPA^{N-terminal mutation}, CEBPA^silenced, together with inv(16), t(15;17), and t(8;21) cases are depicted on the diagonal with a red or blue colored bar. CEBPA^{C-terminal mutation} and CEBPA^{N-terminal mutation} indicates the presence of homozygous mutations. (B) Candidate CEBPA^dm patients and the unambiguous CEBPA^sm patients. The expression levels are defined by the 25-probe set signature. The colors of the hierarchical clustering are relative to the mean. (C) Computed posterior probabilities, indicating the prediction of a CEBPA^dm case, given the 25 predictive probe set signature: P(CEBPA^dm 25-probe sets). The ordering of patients is based on the classification probabilities. (D) True labels (molecular characteristics). Filled lanes indicate the mutation status in CEBPA (CEBPA^dm or CEBPA^sm), NPM1 (NPM1^mutant), or FLT3 (TKD or ITD), open lanes represents no mutation in the particular patient, and a missing value is depicted in gray. *Germline CEBPA^dm cases.