Cell-free DNA size deconvolution resolves nucleosomal origins and reveals tumor-associated fragmentomic alterations

Analysis of cell-free DNA (cfDNA) fragmentomic features holds great promise for minimally invasive cancer diagnostics. Although selectively analyzing short plasma cfDNA enriches tumor-derived DNA (ctDNA), the mechanisms shaping cfDNA size profiles remain incompletely understood. Here, we develop a generalized model of cfDNA fragment length distributions across multiple bodily fluids (saliva, urine, cerebrospinal fluid, lymphatic fluid, and plasma), deconvoluting size profiles into ~10-bp periodic peaks (components), each approximated by a Cauchy-Lorentz distribution. This analytical framework enables investigation of cfDNA fragmentation across diverse pathological states and reveals a 159-bp component that may demarcate intra- and inter-nucleosomal cfDNA. By analyzing plasma DNA from individuals harboring germline TP53 mutations, patients receiving radiotherapy, and liver transplantation recipients, we demonstrate that ctDNA shortening can be distinguished from phagocytosis-associated cfDNA shortening through differences in the amplitude and scale parameters of intra- and inter-nucleosomal components. Moreover, leveraging tumor-related fragmentomic alterations, characterized by increased fragmentation entropy identified through cfDNA size deconvolution, significantly enhances cancer detection.