ENCODE GRAMMAR (ChromBPNet): ChromBPNet models trained on 1,512 DNAse-seq and ATAC-seq experiments across 408 biosamples