Evolutionary and Deep Learning Models Highlight Deleterious Mutations Behind the History of Sugar Beet Breeding

Abstract

Sugar beet (Beta vulgaris ssp. vulgaris) is a major global source of sucrose. Domestication and a history of breeding have shaped the distribution of deleterious mutations across the genome. Accurately resolving these variants offers a compelling avenue to accelerate crop improvement through breeding or genetic engineering. We combine three scales of evolutionary time: protein conservation defined by conservation across all living organisms (SIFT), angiosperm-wide DNA language deep learning model (PlantCaduceus), and fine-scale evolutionary rates from Amaranthaceae plant family sequence alignments to define a high-confidence set of deleterious variants enriched for alleles under negative selection. We then evaluate these deleterious mutations across a multi-study panel of more than 1,900 whole-genome sequenced wild and cultivated beet accessions. We found that domesticated beet accessions exhibit higher deleterious load relative to wild maritima, yet sugar beet carries significantly fewer deleterious variants than other cultivated beet crops. Historical trends show a decline in genetic load across more than a century of sugar beet entries in the U.S. NPGS, reflecting sustained purging during modern breeding. This combinatorial method of leveraging evolutionary conservation reveals how deleterious mutations have shaped sugar beet breeding history and provides potential targets for future selection, causal-variant discovery, and precision genome engineering.