A spatially-aware unsupervised pipeline to identify co-methylation regions in DNA methylation data

DNA methylation data are high-dimensional and spatially structured, which creates substantial multiple-testing burden and can reduce reproducibility in association analyses.

This work introduces SACOMA (Spatially-Aware Clustering for Co-Methylation Analysis), a flexible and unsupervised pipeline that identifies co-methylated regions by combining genomic proximity and methylation similarity through spatially constrained hierarchical clustering.

A tunable, data-adaptive mixing parameter allows SACOMA to avoid rigid threshold assumptions while remaining robust across settings. Although developed for DNAm array studies, the method is generalizable to other spatially dependent data domains.

In simulations and population-level DNAm analyses, SACOMA showed strong sensitivity with effective false-positive control and identified biologically meaningful co-regulated regions, improving both specificity and discovery.