This study presents a multistep approach to identify key drivers of complex traits, such as obesity, by integrating DNA variation and gene expression data with other complex trait data in segregating mouse populations. The method, termed Likelihood-based Causality Model Selection (LCMS), uses conditional correlation measures to determine the best-supported model among three possible relationships between traits: independence, causality, and reactivity. The LCMS procedure was validated using simulated data and an experimental dataset, demonstrating its ability to accurately identify the true model. The approach was then applied to identify genes involved in obesity susceptibility, successfully predicting transcriptional responses to gene perturbations and validating the involvement of three new genes (Zfp90, C3ar1, and Tgfbr2) in obesity. The study highlights the potential of integrating genotypic and expression data to uncover new targets for common human diseases. However, it also discusses limitations, such as measurement errors and the complexity of gene networks, which require further statistical consideration.This study presents a multistep approach to identify key drivers of complex traits, such as obesity, by integrating DNA variation and gene expression data with other complex trait data in segregating mouse populations. The method, termed Likelihood-based Causality Model Selection (LCMS), uses conditional correlation measures to determine the best-supported model among three possible relationships between traits: independence, causality, and reactivity. The LCMS procedure was validated using simulated data and an experimental dataset, demonstrating its ability to accurately identify the true model. The approach was then applied to identify genes involved in obesity susceptibility, successfully predicting transcriptional responses to gene perturbations and validating the involvement of three new genes (Zfp90, C3ar1, and Tgfbr2) in obesity. The study highlights the potential of integrating genotypic and expression data to uncover new targets for common human diseases. However, it also discusses limitations, such as measurement errors and the complexity of gene networks, which require further statistical consideration.