Imputing rare variants in families using a two-stage approach

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Accurate imputation of rare variants is crucial due to increased focus on their study.
• Existing methods for rare variant imputation include reference panel selection and multistage analyses.
• A novel two-stage imputation approach for family data is proposed to integrate strengths of recent advancements.

Purpose of the Study:

• To evaluate a novel two-stage imputation method for enhancing rare variant imputation accuracy in family genetic data.
• To compare the performance of the proposed method against existing imputation tools (Impute2 and Merlin).

Main Methods:

• The study utilized genetic data from chromosome 3 (46.75Mb-49.25Mb) with quality control applied.
• Data included individuals with Genome-Wide Association Study (GWAS) variants only, sequence variants only, or both.
• Prephasing was done using SHAPEIT2; imputation was performed 100 times, masking sequence data for imputation from GWAS data.
• Imputed genotypes (probability > 0.9) were used with minor allele frequency (MAF) cutoffs of 0.01 and 0.05 as input for Merlin, compared to Impute2 and Merlin alone.
• Evaluation metrics included correlation and imputation quality score.

Main Results:

• The proposed two-stage method improved imputation accuracy for variants with MAF between 0.01 and 0.40 when using a MAF cutoff of 0.01 with Impute2 results.
• Accuracy did not improve for variants with MAF < 0.01 under the same MAF cutoff.
• A MAF cutoff of 0.05 resulted in the proposed two-stage approach performing worse than other methods for variants with MAF between 0.01 and 0.40.

Conclusions:

• The two-stage imputation method shows potential but requires optimization, possibly by adjusting Impute2 variant inclusion criteria.
• Further research on larger genomic regions and varied inclusion thresholds is necessary to fully assess the method's accuracy.
• The findings highlight the sensitivity of imputation accuracy to MAF cutoffs and method parameters.