Incorporating Sex Differences Improves Genomic Prediction of Food Intake Behavior in <em>Drosophila melanogaster </em>

Insects exhibit remarkable evolutionary success due to their rapid adaptation and resilience to environmental changes. Advances in high-throughput sequencing have expanded our ability to predict quantitative trait phenotypes using high-resolution genomic polymorphism data. This study assesses the predictive accuracy of two statistical models&mdash;GBLUP and Bayes B&mdash;for food intake traits in Drosophila melanogaster, leveraging ~1.96 million SNPs from the Drosophila Genetic Reference Panel of inbred lines. We explored whether prediction accuracy varies by trait and sex, analyzing male and female phenotypes independently. Using 5-fold cross-validation, we measured the predictive ability of a model as the correlation between predicted genetic values and observed phenotypes. Predictive accuracies for the food intake trait were 0.0368 &plusmn; 0.0103 and 0.0687 &plusmn; 0.0203 for GBLUP, and 0.0329 &plusmn; 0.0379 and 0.0239 &plusmn; 0.0138 for Bayes B, in females and males, respectively. These results reveal that genetic architecture significantly influences prediction outcomes, with trait complexity and sex-specific genetic effects shaping model performance. Notably, differences in accuracy across sexes underscore the need for tailored statistical approaches in genomic selection. Our findings enhance the understanding of genomic prediction in studying local adaptations and evolutionary dynamics in fruit flies, offering broader implications for decoding phenotypic variation in genetically diverse species.