Datasets for chromatin hub prediction in six cell lines based on multiple genomic features

Tables with features and classes for machine learning prediction of chromatin hubs. Genomic features include CTCF, EP300, H3K27me3, H3K36me3, H3K4me1, H3K4me2, H3K4me3, H3K9ac, H3K9me3, RAD21, RNAPol2, and RNA.Seq, while the classes are Hubs and Non-Hubs.

The cell lines featured here are A549, H1ESC, HeLa, IMR90, K562, and MCF7. They happen to be the 6 cell lines out of 8 existing in our integrative database, GREG (https://doi.org/10.1093/database/baz162). The normalized read-coverages from features (variables) are mapped through genomic intervals of 2 Kbs, genome-wide. Such genomic intervals (bins), are classified as Hubs or Non-Hubs. Hubs are those bins with multiple chromatin interactions, including at least one long-range interaction (larger than 1Mb) or an inter-chromosomal interaction (tagged as Inf).

Columns per table:
chr    start    end    CTCF    EP300    H3K27me3    H3K36me3    H3K4me1    H3K4me2    H3K4me3    H3K9ac    H3K9me3    RAD21    RNA.Seq    RNAPol2    Class

Note that features may be inconsistent across different cell types, due to the availability of data. The BAM files have been sourced from ENCODE and NCBI repositories.

The analysis following this data can be found at https://github.com/mora-lab/GREG-Hubs.