Iteratively reweighted LASSO for mapping multiple quantitative trait loci.

Area of Science:

• Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Iteratively Reweighted Least Square (IRLS) and Maximum Likelihood (ML) are common for quantitative trait locus (QTL) detection.
• IRLS offers faster computation and robust parameter estimation over ML.
• However, IRLS has limited statistical power for detecting multiple QTLs under a single QTL model.

Purpose of the Study:

• To introduce the Iteratively Reweighted Least Absolute Shrinkage and Selection Operator (IRLASSO) method.
• To extend IRLS capabilities for simultaneous multiple QTL mapping.
• To enhance the precision of parameter estimation and QTL detection power.

Main Methods:

• Proposed IRLASSO method integrating IRLS with the Least Absolute Shrinkage and Selection Operator (LASSO).
• Utilized LASSO with a coordinate descent step for efficient estimation of genetic effects across the genome.
• Applied the method to a barley dataset from the North American Barley Genome Mapping Project.

Main Results:

• IRLASSO demonstrated higher precision in parameter estimation and increased power for QTL detection compared to IRLS.
• IRLASSO accurately estimated residual variance, outperforming the unweighted LASSO based on Ordinary Least Squares (LS).
• The method exhibited high computational efficiency, typically converging within five iterations.

Conclusions:

• IRLASSO effectively addresses the limitations of IRLS in detecting multiple QTLs.
• The proposed method offers a computationally efficient and statistically powerful approach for multiple QTL analysis.
• IRLASSO provides a valuable tool for genetic studies, as demonstrated by its application to barley data.