Data preparation

The raw and processed NB and GCB datasets were downloaded from the Gene Expression Omnibus (GEO), under accession numbers {"type":"entrez-geo","attrs":{"text":"GSE84022","term_id":"84022"}}GSE84022 (for RNA-seq FASTQ files), {"type":"entrez-geo","attrs":{"text":"GSE159314","term_id":"159314"}}GSE159314 (for ATAC-seq narrow peaks) and {"type":"entrez-geo","attrs":{"text":"GSE84022","term_id":"84022"}}GSE84022 (for Hi-C contacts). Data relating to the narrow peaks of the six histone marks H3K4me1, H3K4me3, H3K9me3, H3K27me3, H3K36me3 and H3K27ac were downloaded from DeepBlue (16). The RNA-seq FASTQ files generated using the following cancer cell lines were obtained from the Cancer Cell Line Encyclopedia (17): OCI-LY10 (Ly10), activated B-cell-like (ABC) subtype of diffuse large B-cell lymphoma (DLBCL); OCI-LY7 (Ly7), GCB-cell subtype of DLBCL; BL-41, Burkitt lymphoma; and Mino, mantle cell lymphoma. Data relating to the ChIP-seq peaks of BACH2 ({"type":"entrez-geo","attrs":{"text":"GSM1084800","term_id":"1084800"}}GSM1084800), MAFK ({"type":"entrez-geo","attrs":{"text":"GSM1159670","term_id":"1159670"}}GSM1159670) and MYC ({"type":"entrez-geo","attrs":{"text":"GSM762710","term_id":"762710"}}GSM762710) were also downloaded from GEO.

The enhancer–promoter interactions (EPIs) screened by the CRISPRi-FlowFISH method for chronic myeloid leukemia (CML) cell line K562 were downloaded from the original study (11). The EPIs predicted by the ABC model (11) for acute monocytic leukemia (AML) cell line THP-1 and OCI-LY7 were downloaded from the published research (18). The gene expression profiles for K562 and THP-1 were downloaded from DepMap (https://depmap.org/portal/), and the histone modifications and chromatin accessibility datasets were prepared from ENCODE, BLUEPRINT and GEO, as summarized in Supplementary Table S1.