Imputation of low-frequency variants using the HapMap3 benefits from large, diverse reference sets.

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Genotype imputation infers unobserved genetic data in low-density datasets.
• It is crucial for disease association studies, particularly for variants not well-covered by standard genotyping arrays, such as low-frequency variants.
• The impact of reference panel choice on imputation accuracy is less understood than algorithm development.

Purpose of the Study:

• To evaluate the impact of reference panel size and diversity on genotype imputation accuracy.
• To compare the performance of HapMap2 versus the newer HapMap3 reference panels.
• To assess improvements for low-frequency variants specifically.

Main Methods:

• Comparative analysis of imputation accuracy using HapMap2 and HapMap3 reference panels.
• Assessment of imputation quality scores across different reference set sizes and population compositions.
• Focus on imputation into Western European samples.

Main Results:

• HapMap3 provides more accurate imputation and better-calibrated quality scores than HapMap2 for Western European samples.
• Increasing the number of HapMap3 populations in the reference set further enhances imputation accuracy.
• Improvements are most significant for low-frequency variants (<5%), with large, diverse panels nearing the accuracy of common variants.
• Reference set diversity improves low-frequency variant imputation accuracy independently of sample size.

Conclusions:

• HapMap3 reference panels offer substantial improvements in imputation accuracy over HapMap2.
• These gains are particularly valuable for accurate imputation of low-frequency variants.
• Future large-scale projects like the main 1000 Genomes Project are expected to enable reliable imputation of low-frequency variants.