On the differences between maximum likelihood and regression interval mapping in the analysis of quantitative trait...

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Quantitative trait loci (QTL) analysis is crucial for understanding complex traits.
• Maximum-likelihood (ML) and regression (REG) are common interval mapping methods.
• Understanding their differences is key for accurate genetic analysis.

Purpose of the Study:

• To analytically and numerically investigate the differences between ML and REG interval mapping for QTL analysis.
• To identify factors influencing the discrepancies between these two methods.
• To provide insights for efficient QTL mapping strategies.

Main Methods:

• Analytical comparison of ML and REG solution sets for QTL parameter estimation.
• Numerical simulations to evaluate differences in Mean Squared Error (MSE), Likelihood-Ratio Test (LRT) statistics, and QTL detection power.
• Investigation of factors like variance explained, QTL position, interval size, epistasis, and linkage.

Main Results:

• ML and REG differences are influenced by variance explained, QTL position, interval size, epistasis, and linkage.
• REG method shows bias in estimating variance explained and struggles with closely linked QTL.
• ML method generally provides more powerful, accurate, and precise estimates with smaller MSEs and larger LRT statistics.
• Differences are more pronounced with higher variance explained, central QTL positions, wider intervals, stronger epistasis, larger QTL effect differences, and closer QTL positions.

Conclusions:

• ML interval mapping is generally more accurate, precise, and powerful than REG interval mapping.
• REG is computationally faster, especially for large models.
• Recognizing factors influencing method differences aids in developing efficient, combined QTL mapping strategies.