DSRIG: Incorporating graphical structure in the regularized modeling of SNP data

Area of Science:

• Animal Genetics
• Statistical Genomics
• Quantitative Genetics

Background:

• Genetic selection is crucial for farm animal improvement.
• Regularized regression on single nucleotide polymorphism (SNP) data aids in identifying genes associated with desired traits.
• Accounting for SNP relationships is vital to avoid undesirable breeding outcomes.

Purpose of the Study:

• To introduce Doubly Sparse Regression Incorporating Graphical structure (DSRIG), a novel regularized method for genetic selection.
• To leverage relationships among candidate SNPs for improved prediction accuracy.
• To identify SNPs associated with boar taint compounds (skatole and androstenone).

Main Methods:

• Application of DSRIG to predict skatole and androstenone levels.
• Comparison of DSRIG with Ordinary Least Squares (OLS) and Least Absolute Shrinkage and Selection Operator (LASSO) using cross-validation.
• Analysis of coefficient estimates to determine SNP impact.
• Use of a consensus graph to infer SNP relationships.

Main Results:

• DSRIG demonstrated a predictive benefit over OLS and LASSO in cross-validation.
• The method identified SNPs with potentially significant impacts on boar taint compound levels.
• Inferred relationships among SNPs provide insights into genetic architecture.

Conclusions:

• DSRIG offers an improved approach for genetic selection by incorporating SNP network structures.
• The method effectively predicts complex traits influenced by multiple genetic factors.
• Findings contribute to more precise genetic improvement programs in livestock.