Incorporating genotyping uncertainty in haplotype inference for single-nucleotide polymorphisms

Area of Science:

• Genetics and Genomics
• Bioinformatics
• Statistical Genetics

Background:

• High-throughput genotyping generates vast genotypic data essential for complex disease research.
• Current automated programs often lack quality control for allele calls, necessitating manual, error-prone interventions.
• Accurate genotype data is critical for reliable haplotype analyses and linkage-disequilibrium mapping.

Purpose of the Study:

• To introduce a novel, automated genotype clustering algorithm, GeneScore, for improved allele call accuracy.
• To present an expectation-maximization (EM) algorithm, GenoSpectrum (GS)-EM, for probabilistic haplotype phasing.
• To demonstrate the combined efficacy of GeneScore and GS-EM for direct haplotype inference from raw genotyping data.

Main Methods:

• GeneScore: A novel genotype clustering algorithm utilizing a bivariate t-mixture model.
• GenoSpectrum (GS)-EM: An expectation-maximization (EM) algorithm for haplotype phasing using probabilistic genotype matrices.
• Integration of GeneScore and GS-EM for direct haplotype inference from raw genotyping machine outputs (e.g., TaqMan assay).

Main Results:

• The probabilistic approach significantly enhances the accuracy of genotype calling and haplotype inference compared to existing methods.
• Simulated and real data analyses confirm the superior performance of GeneScore and GS-EM.
• Improved accuracy leads to increased statistical power in haplotype-based association analyses for complex diseases.

Conclusions:

• GeneScore and GS-EM provide a robust, model-based framework for accurate genotype and haplotype inference.
• This probabilistic approach reduces reliance on manual quality control, saving labor and minimizing errors.
• The method offers enhanced accuracy and statistical power, advancing genetic studies of complex diseases.