Fig. 2: Identification of intratumor malignant meta-programs in PCNS DLBCL patients.

a Heatmap depicting pair-wise correlations of intratumor programs derived from PCNS DLBCL patients. Both rows and columns represent programs identified by consensus NMF algorithm sample-wise. Color bar on the left indicates which sample each program belongs to. Color of each block represents the Pearson correlation coefficient (PCC) between two programs. Further hierarchical clustering identified seven coherent expression meta-programs across samples (MP1–7). The representative genes of each meta-program are displayed on the right of the heatmap. b Heatmap of representative biological pathways significantly enriched in the seven meta-programs. UMAP of malignant cells colored by datasets (c), cancer types (d), samples (e), and COO classification (f). FL follicular lymphoma, tFL transformed follicular lymphoma, DLBCL diffuse large B-cell lymphoma. g Boxplot showing the proportion of cells enriched for the meta-programs in each patient. Each dot in the boxplot represents the cell proportion of an individual patient, and patients were classified into different cancer types. A two-sided Wilcoxon rank-sum statistic was used to calculate significance (***P < 0.001, **P < 0.01, *P < 0.05). Box boundaries and middle lines correspond to the interquartile range (IQR) and median, respectively. Whiskers extend to the lowest or highest data points that are no more than 1.5 times the IQR from the box boundaries. In (c), Steen et al.¹⁴, Zhang et al.²⁷ and Roider et al.¹³ represent three publicly available scRNA-seq datasets; “pcnsl” represents PCNS DLBCL dataset in this study.