CvN method: identification of the nearest normal population for individual AK-AML samples, using a gene expression-based landscape of the normal hematopoietic hierarchy. (A) PCA of gene expression profiles from the following normal purified BM populations: HSCs, multipotent progenitors (MPPs), common myeloid progenitors (CMPs), granulocyte-monocyte progenitors (GMPs), megakaryocyte-erythrocyte progenitors (MEPs), early PM, late PM, myelocytes (MY), metamyelocytes (MM), band cells (BC), polymorphonuclear neutrophilic granulocytes (PMN_BM), and monocytes (Mono). The PCA was performed on 2119 probe sets (1367 genes) selected by variance filtering. (B) Spearman correlation matrix of gene expression of the samples from A. (C) Workflow of the CvN method for the identification of the nearest normal counterpart for individual AK-AML samples (CvN method): AML samples are normalized individually together with the data set of the normal hematopoietic hierarchy shown in A and B, and the normal populations closest to AML samples are identified by a 2-step approach. First, the Euclidian distances between each individual AML sample and all the normal blood and BM populations are calculated using gene expression profiles projected onto the first 6 principal components. Next, the 50% most varying probe sets within the 15 closest normal populations are selected for each individual AML samples and used in a second PCA to map the AML sample to its 5 nearest normal populations. Subsequently, a weighted average gene expression profile based on the Euclidian distance between the 5 normal populations and AML sample is calculated. Finally, gene expression profiles of AML samples and their corresponding individual average-weighted normal population are compared with defined differentially expressed genes in individual AML samples for enrichment analysis, prognostification, and further analyses.