ReadsRatio: a facile method for cell type deconvolution with cross-platform applications

Existing methylation-based cfDNA deconvolution methods are sensitive to sequencing depth, reference quality, and platform variability. We develope ReadsRatio, an approach to estimate tissue proportions using the ratio of unmethylated reads within cell type-specific marker regions, without requiring reference methylation intensity signals. Compared to conventional methods, ReadsRatio detects specific tissue involvement in both simulations and clinical cohorts involving lupus, kidney transplantation, and COVID-19. It maintains accuracy across sequencing platforms and at reduced sequencing depths, enabling substantial cost savings in our preliminary estimations. ReadsRatio provides a versatile, cost-efficient framework for cfDNA deconvolution with broad research and clinical applicability.