Polymer-derived distance penalties improve chromatin interaction predictions from single-cell data across crop genomes

Scalable proxies of 3D genome interactions, such as from single-cell co-accessibility or Deep Learning, systematically overestimate long-range chromatin contacts. To correct this bias, we introduce a penalty function grounded in polymer physics, derived by fitting a multi-component power-law model to experimental Hi-C data from maize, rice, and soybean. This correction substantially improves concordance with Hi-C, reduces false-positive rates of long-range interactions by up to 95%, and reveals distinct decay exponents corresponding to different scales of chromatin organization. We provide open-source code and derived parameters to facilitate broad application across plant species.