Estimation of allele frequency and association mapping using next-generation sequencing data

Area of Science:

• Population genetics
• Genomic data analysis
• Bioinformatics

Background:

• Accurate allele frequency estimation is crucial for population genetics and association mapping.
• Low-coverage next-generation sequencing (NGS) data (<15X) is cost-effective but presents challenges due to high error rates and coverage variability.
• Existing SNP calling methods struggle with statistical uncertainty in low-coverage data.

Purpose of the Study:

• To evaluate a novel maximum likelihood (ML) method for estimating allele frequencies using low and medium-coverage NGS data.
• To compare the performance of the ML method against traditional genotype calling methods.
• To assess the utility of the ML method for direct association testing in case/control studies.

Main Methods:

• Developed and applied a maximum likelihood method that integrates over individual genotype uncertainty, avoiding preliminary genotype calling.
• Utilized simulations to compare the ML method with genotype calling-based approaches.
• Validated findings using real re-sequencing data from 200 individuals from an exon-capture experiment.

Main Results:

• The ML method demonstrated superior accuracy in allele frequency estimation compared to genotype calling methods.
• The ML method provided more accurate estimations of allele frequency distributions across neutrally evolving sites.
• Association mapping studies using the ML method showed increased statistical power.
• Simulation results were corroborated by analyses of real NGS data.

Conclusions:

• Genotype calling should be avoided for association mapping and allele frequency estimation with low to medium-coverage NGS data.
• Filtering genotypes based on call confidence scores can be detrimental when using genotype calling methods.
• The proposed ML method offers a more robust alternative for analyzing low-coverage sequencing data.