Improving accuracy of rare variant imputation with a two-step imputation approach

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Genome-wide association studies (GWAS) rely on genotype imputation for variant discovery.
• Traditional imputation methods using limited reference panels (e.g., HapMap) underrepresent low-frequency and rare variants (MAF <5%).
• Advancements in genotyping arrays and reference panels (e.g., 1000 Genomes) aim to improve imputation of these less frequent variants.

Purpose of the Study:

• To present a novel two-step genotype imputation approach.
• To enhance imputation quality for low-frequency single-nucleotide polymorphisms (SNPs).
• To offer a cost-effective strategy for large-scale epidemiological studies.

Main Methods:

• Developed a two-step imputation strategy.
• Created a local reference population using a subset of samples genotyped on a dense array.
• Imputed study samples first to the local reference, then to the 1000 Genomes panel.
• Assessed imputation quality using r-squared and concordance rates.

Main Results:

• The two-step approach increased mean imputation quality by 28% for variants with MAF between 1% and 5% compared to direct imputation.
• Significantly higher concordance rates were observed for heterozygotes and rare homozygotes in the low-frequency range (P<1e-15).
• The method demonstrated improved imputation accuracy in the low-frequency spectrum.

Conclusions:

• The proposed two-step imputation method significantly improves the quality of genotype imputation for low-frequency variants.
• This approach is a cost-effective enhancement over traditional direct imputation for large epidemiological studies.
• The strategy facilitates more comprehensive scrutiny of less frequent variants in GWAS.