Variable Genetic Architecture Limits Cross‐Range Genomic Prediction of Functional Traits in Invasive Common Ragweed ( <i>Ambrosia artemisiifolia</i> )

ABSTRACT Functional trait expression is a potentially powerful predictor of invasion success, but the reliability of trait‐based predictions across populations depends on the consistency of genetic architecture. We tested this assumption using common ragweed ( Ambrosia artemisiifolia ), an annual plant native to North America that has invaded Europe and Australia. Using genome‐wide markers (158,652 SNPs) and five functional traits measured in a common garden, we characterised trait correlations, genetic architecture, evolutionary potential, and the transferability of genomic prediction (GP) models across the native and invasive ranges. Genetic and phenotypic diversity and their correlation declined from the native range through successively more recent invasive populations, with Australian plants showing the lowest diversity and weakest phenotypic integration. GP models trained using genotypes from one range were consistently poor at predicting trait values in other ranges, indicating that the genetic architecture underlying functional traits is not conserved across ranges. Prediction accuracy improved substantially when significantly correlated traits were included as additional predictors, but only where the strength and direction of those correlations were conserved, which was not the case between all ranges. Multi‐trait analyses of evolutionary potential revealed declining phenotypic integration in invasive ranges, suggesting these populations have not yet reached trait optima. Our results highlight how the demographic and evolutionary processes inherent to biological invasions reshape genetic architecture and trait relationships, limiting the use of conventional genomic prediction models and instead requiring integrated frameworks that account for species biology to predict invasion outcomes.