Improving stability and understandability of genotype-phenotype mapping in Saccharomyces using regularized variable...

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Multivariate approaches are vital for genome-wide association studies (GWAS).
• Previous Partial Least Squares (PLS) methods mapped yeast genotype-phenotype relations using background information post hoc.
• L-Partial Least Squares (L-PLS) integrates background information during the modeling phase.

Purpose of the Study:

• To introduce a modified L-PLS approach with Variable Importance on Projection (VIP) for enhanced genotype-phenotype mapping in GWAS.
• To evaluate the effectiveness of integrating background information at the modeling level.
• To compare the modified L-PLS method with traditional PLS procedures.

Main Methods:

• Implementation of a modified L-Partial Least Squares (L-PLS) with Variable Importance on Projection (VIP).
• Utilized a stepwise regularized procedure for selecting both genes and background information.
• Compared results against PLS-based procedures that did not incorporate background information.

Main Results:

• The modified L-PLS approach improved the understandability of genotype-phenotype relationships in yeast.
• Phenotypic variations were explained by stable genes and background variations.
• Phenotype-influencing genes showed a 29% faster evolution rate compared to non-influential genes, consistent with other studies.

Conclusions:

• A modified L-PLS with VIP in a stepwise regularized elimination procedure enhances the understandability and stability of selected genes and background information.
• This approach is recommended for genome-wide association studies (GWAS) when background information is available.
• The methodology offers an automated solution for genotype-phenotype mapping.