A two-phase procedure for QTL mapping with regression models.

Area of Science:

• Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Quantitative trait loci (QTL) mapping often involves a large number of genetic markers.
• Standard regression models face challenges when the number of features exceeds the sample size, particularly when considering epistasis.
• Conventional stepwise procedures exhibit greedy behavior, which is exacerbated in high-dimensional genetic data.

Purpose of the Study:

• To propose a robust two-phase procedure for QTL mapping that addresses the challenges of high-dimensional genetic data.
• To enhance the accuracy and efficiency of identifying quantitative trait loci (QTL) by incorporating penalized likelihood and extended Bayes information criterion (EBIC).

Main Methods:

• A two-phase approach combining penalized likelihood techniques and extended Bayes information criterion (EBIC).
• Phase 1: Screening of main and interaction features using a penalized likelihood mechanism.
• Phase 2: Low-dimensional selection using EBIC on retained features to identify QTL.

Main Results:

• The proposed two-phase procedure demonstrates asymptotic properties where the positive detection rate (PDR) converges to 1 and the false discovery rate (FDR) converges to 0 as sample size increases.
• Simulation studies show competitive performance compared to traditional and recent QTL mapping methods.
• A real data analysis validates the practical applicability of the procedure.

Conclusions:

• The developed two-phase procedure offers an effective solution for QTL mapping in high-dimensional genetic datasets.
• The method provides reliable identification of quantitative trait loci (QTL) with controlled false discoveries.
• This approach advances statistical genomics by offering a powerful tool for genetic marker analysis.