Structural variation landscape reveals phenotypic divergence across cocoa populations

Abstract

Abstract Structural variations (SVs) represent an important source of genomic diversity and can contribute substantially to phenotypic variation in crops. However, the population-scale distribution and phenotypic effects of SVs in Theobroma cacao L. (cocoa) remain poorly understood. Here, we constructed a population-scale cocoa SV atlas using whole-genome resequencing data from 165 cocoa accessions representing ten previously defined genetic groups. Using a unified short-read-based SV detection strategy, we identified 11 271 high-confidence SVs, including deletions, duplications, and inversions. Using a framework based on term frequency-inverse document frequency (TF-IDF) algorithm, we identified 1078 fingerprint SVs for ten cocoa genetic groups of which 145 showed significant SV–trait associations. To further investigate integrated phenotypic divergence associated with functional SVs, we developed an analysis framework based on latent Dirichlet allocation (LDA). This analysis identified four latent phenotypic features showing significant divergence among cocoa genetic groups. Geographic populations from South America displayed extensive admixture of genetic groups, whereas populations outside South America showed reduced genetic diversity consistent with historical dispersal bottlenecks. Our study provides a population-scale SV resource for cocoa and demonstrates that SVs contribute to genetic differentiation, phenotypic divergence, and geographic adaptation in cocoa populations.