Haplotype analysis in the presence of informatively missing genotype data

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Missing genotype data is prevalent in genetic studies.
• Existing methods often assume missing data is random, which may not be true.
• Violating this assumption can lead to biased haplotype frequency estimates and inaccurate association analyses.

Purpose of the Study:

• To develop a general missing data model for genetic markers.
• To address biases caused by the missing at random assumption in haplotype analysis.
• To improve the accuracy of haplotype frequency estimation and association testing.

Main Methods:

• Proposed a general missing data model for multiple markers.
• Proved identifiability conditions for haplotype frequencies and missing data probabilities.
• Conducted simulation studies using two single nucleotide polymorphisms (SNPs).

Main Results:

• Violation of the missing at random assumption causes significant bias.
• The proposed model reduces bias in haplotype frequency estimates.
• The new model improves accuracy in haplotype association analyses, reducing false positives and negatives.

Conclusions:

• The proposed general missing data model effectively handles unknown missing data mechanisms.
• This approach is crucial for accurate genetic association studies with missing genotype data.
• The method demonstrated utility in a real-world genetic dataset.