Rare variant analysis reveals divergent genetic architectures of directly selected and correlated silkworm traits

Correlational selection acts on combinations of correlated traits, driving multivariate phenotypic evolution, yet how it differentially shapes direct versus correlated traits remains unclear. Through rare variant association analysis of silkworm silk-yield traits, we reveal divergent genetic architectures of the direct traits cocoon shell weight (CSW) and ratio (CSR) versus the correlated traits cocoon and pupal weight (CW and PW). Direct targets were governed by fewer genes (11 and 12) enriched for deleterious, high-impact, negative-effect rare variants arising de novo during domestication and purged during improvement. Correlated traits involved more genes (25 and 65) with positive effects and moderate-impact variants largely inherited from wild silkworm, with frequency declines reflecting neutral processes. Selection signals confirmed stronger selection on CSW/CSR-associated genes. G-matrix analysis also revealed strong positive genetic correlations ( rG : 0.336–0.814) during domestication, partly decoupled during improvement. Our study elucidates how artificial correlational selection altered historical correlations between these traits and shaped their distinct genetic architectures, providing empirical insights into correlational selection.