This dataset contains pre-compiled reference sets for use with the CoRE-BED model that note genome-wide status of 19 genomic features (TSS proximity, H3K27ac, H3K27me3, H3K4me1, H3K4me2, H3K4me3, H3K36me3, H3K79me2, H3K9ac, H3K9me3, H4K20me1, ATAC-seq, DNase-seq, CTCF, EP300, H2AFZ, POL2RA, RAD21, and SMC3) across 816 biosamples encompassing 33 major cell and tissue types. Also included are the raw peak calls in UCSC BED format that were used to generate these reference files. Additionally, we include genome-wide sets of functional predictions generated using the CoRE-BED framework across all 816 biosamples, the discrete combinations of 19 genomic features found in each of the major cell and tissue types, and the datasets used to train the CoRE-BED model. Relevant data are made available with both hg19 and hg38 genome build alignments. Please cite: Betti, M.J., Aldrich, M.C., & Gamazon, E.R. (2023). Maximum-entropy integrative epigenomic framework reveals regulatory mechanisms at human disease loci. Preprint. Betti, M.J., Aldrich, M.C., & Gamazon, E.R. (2023). CoRE-BED datasets (Version 2). Zenodo. 10.5281/zenodo.7558115