Decoding the sequence determinants of locus-specific DNA methylation across human tissues

Researchers developed Melody, a deep learning model that predicts DNA methylation patterns from 10-kb genomic sequences across 39 human tissues, outperforming existing methods. Melody also predicts methylation quantitative trait locus effects and identifies regulatory motifs. An extension, Melody-G, uses single-cell RNA-seq data to infer methylation in unprofiled cell types, linking genomic sequence to cellular methylation states.