Handling missing values in population data: consequences for maximum likelihood estimation of haplotype frequencies

Area of Science:

• Population Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Haplotype frequency estimation is crucial in population genetics.
• Expectation-maximization (EM) methods are widely used but primarily assessed for complete datasets.
• Missing genotypes are common in large-scale genetic studies.

Purpose of the Study:

• To develop and evaluate an extended EM method for haplotype frequency estimation with missing genotypes.
• To compare the performance of the proposed method against common strategies for handling missing data.
• To provide guidelines for managing missing genotype data in genetic analyses.

Main Methods:

• An extension of the EM algorithm was developed to accommodate missing genotypes.
• Simulations were conducted using hematopoietic stem cell donor data genotyped at three HLA loci.
• Missing genotypes were artificially introduced at various proportions.
• The proposed method was compared with methods that ignore incomplete data or treat missing alleles generically.

Main Results:

• The proposed EM extension demonstrated superior qualitative and quantitative haplotype frequency estimations.
• The method's performance was evaluated against complete data benchmarks.
• Increased computation time was observed with the proposed method.

Conclusions:

• The developed EM method effectively handles missing genotypes for accurate haplotype frequency estimation.
• Missing data significantly impacts estimation accuracy, necessitating robust handling strategies.
• Guidelines are proposed for routine missing data management in population genetics studies.