When Does Choice of Accuracy Measure Alter Imputation Accuracy Assessments?

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genetics

Background:

• Genotype imputation refines genetic association findings but inaccuracies can distort results.
• Existing accuracy metrics may not adequately assess imputation quality, particularly for rare variants.

Purpose of the Study:

• To compare the Imputation Quality Score (IQS) with other common accuracy assessment statistics.
• To evaluate the performance of different imputation accuracy metrics across variant frequencies.

Main Methods:

• Compared IQS with concordance rate, squared correlation, and imputation program-specific metrics (BEAGLE R2, IMPUTE2 INFO).
• Used 1000 Genomes reference panels and sequenced subjects from nicotine dependence studies for imputation.
• Masked and imputed genotypes using BEAGLE 3.3.2 and IMPUTE2.

Main Results:

• Concordance rate inflates accuracy estimates, especially for rare and low-frequency variants.
• For common variants, IQS, squared correlation, BEAGLE R2, and IMPUTE2 INFO show similar results.
• For rare and low-frequency variants, IQS is more conservative than other metrics.

Conclusions:

• IQS offers a more reliable assessment of imputation accuracy, particularly for rare and low-frequency variants.
• Accurate imputation quality assessment is vital for robust genetic association studies.