Fig. 2

Distinct ncRNA expression profiles characterize cells of the different human blood lineages. All depicted data refer to the Arraystar Human lncRNA Microarray V2.0 platform. a–c t-SNE of all samples using the most variable a mRNAs (3926), b ncRNAs (3151), and c lincRNAs (767). d Self-organizing maps (SOMs) trained using the 17,655 most variable mRNAs and ncRNAs in 11 sample groups. Black rectangles: group-specific overexpression spots. Center: neighbor-joining tree built using the 68 lineage-specific spot metagenes. e Heatmaps of (left) 2493 fingerprint ncRNAs and (right) 581 anti-fingerprint ncRNAs, defined by integrating SOM and limma analyses. f Guilt-by-association workflow for the fingerprint/anti-fingerprint ncRNAs and all protein-coding genes. g Enrichment map network analysis for HOTAIRM1 (FDR < 0.05, see the methods section for details). Circle size corresponds to the size of the gene set, and connecting line thickness represents the degree of similarity between two gene sets. Red and blue nodes indicate positive and negative correlation to HOTAIRM1 expression, respectively. Gene set labels printed in bold indicate a similar association (FDR < 0.05) observed in at least one AML validation cohort