Identifying critical causal SNPs for trait analysis in alfalfa (Medicago sativa L.) via a causal graphical framework

Authors: LEE, YANGMING - Rochester Institute Of Technology; MEDINA, CESAR - University Of Minnesota; Xu, Zhanyou

Submitted to: Scientific Reports
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/8/2026
Publication Date: N/A
Citation: N/A

Interpretive Summary: Alfalfa is one of the most important forage crops in the United States, supporting the dairy industry and improving soil health. However, improving traits like winter survival, stem quality, and yield has been challenging due to the crop’s complex genetics. To address this, researchers developed a new method that goes beyond traditional genetic studies. Instead of just identifying which genes are associated with traits, this new approach uncovers which genes actually cause those traits. By using a “causal graphical framework,” the team was able to pinpoint a small number of genetic markers that directly influence key traits in alfalfa. This breakthrough helps plant breeders focus on the most important genes, making breeding more efficient and precise. Ultimately, it could lead to stronger, more productive alfalfa varieties that better withstand harsh conditions—benefiting farmers, the environment, and the food system as a whole.

Technical Abstract: Alfalfa (Medicago sativa), as a vital forage crop, urgently requires optimization of different agronomic traits to improve total biomass, forage quality, and tolerance to extreme environments. Traditional approaches, such as Genome-Wide Association Studies (GWAS) and machine learning methods, have identified statistical associations between single nucleotide polymorphisms (SNPs) and various traits. However, these techniques cannot distinguish SNPs that have a critical impact on traits from those that are only indirectly associated, which hinders our understanding of aggregate genetic mechanisms and constrains advances in breeding efficiency and accuracy. This study presents a novel Causal Graphical (CG) Framework for alfalfa stem traits, aiming to reveal causal chains of influence among SNPs. The method begins with feature association analysis and trait-based feature importance evaluation to independently assess SNP relevance, followed by a joint down-selection of candidate SNPs for causal modeling. For causal discovery, a constraint-based method is employed to identify potential causal links between SNPs and traits, which are subsequently validated using a score-based method to eliminate features with indirect impacts on traits. The resulting causal graphs were compared with SNPs identified by Random Forest (RF) and Support Vector Machine (SVM), and the results demonstrated that this framework pinpoints a more focused set of 137-203 putative causal SNPs per trait, including only 4-5 direct causal drivers, which are also commonly recognized by both RF and SVM.