Decision tree resulting from recursive partitioning analysis and amalgamation in the training series. Disruption of TP53 and BIRC3, mutations of SF3B1 and NOTCH1, and del11q22-q23 were the factors selected by the algorithm to split the patient population in 6 terminal nodes. Presence or absence of the TP53 disruption independent of cooccurring genetic lesions was the most significant covariate for the entire study population. Among patients lacking TP53 abnormalities, the most significant covariate was BIRC3 disruption. Among patients lacking both TP53 and BIRC3 abnormalities, the most significant covariate was SF3B1 mutation status. Among patients lacking TP53, BIRC3, and SF3B1 lesions, the most significant covariate was NOTCH1 mutation status. Among patients lacking TP53, BIRC3, SF3B1, and NOTCH1 lesions, the most significant covariate was del11q22-q23. Based on the application of the amalgamation algorithm to the terminal nodes, patients harboring TP53 abnormalities and those harboring BIRC3 abnormalities were grouped into a single category, as well as patients harboring NOTCH1 mutations, SF3B1 mutations, or del11q22-q23. Genetic lesions are represented from right to left according to their hierarchical order of relevance in splitting the parent node into daughter nodes with significantly different survival probabilities. The P value corresponds to the log-rank test adjusted for multiple comparisons. The right branch of each split represents the presence of the lesion. The left branch of each split represents the absence of the lesion. The Kaplan-Meier curves estimate the OS of patients belonging to each terminal node. N indicates the number of patients in the node; M, mutation; and DIS, disruption.