A logistic regression mixture model for interval mapping of genetic trait loci affecting binary phenotypes.

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Genetic studies often involve complex traits influenced by multiple factors, including covariates.
• Detecting specific genetic loci (binary trait loci or BTLs) is crucial for understanding disease inheritance.
• Existing methods may not adequately account for covariates in BTL detection.

Purpose of the Study:

• To develop and evaluate finite logistic regression mixture models for BTL detection.
• To establish accurate statistical thresholds for identifying BTLs using interval mapping.
• To compare the performance of proposed methods against traditional approaches.

Main Methods:

• Utilized finite logistic regression mixture models for interval mapping.
• Derived maximum-likelihood estimates (MLEs) for logistic regression parameters.
• Determined null asymptotic distributions of likelihood-ratio test (LRT) statistics using a chi2-process.
• Employed Monte Carlo methods to approximate thresholds for BTL detection.

Main Results:

• MLEs for logistic regression parameters were found to be asymptotically unbiased.
• Null asymptotic distributions of LRT statistics were explicitly determined.
• Accurate thresholds derived from limiting distributions significantly reduce false BTL detection rates.
• Proposed methods demonstrate good performance in moderately large sample sizes.

Conclusions:

• Finite logistic regression mixture models provide a robust framework for BTL detection.
• Accurate threshold determination is critical to avoid excessive false positive rates.
• The developed methods offer improved accuracy and reliability in genetic association studies.