How local reference panels improve imputation in French populations

Area of Science:

• Genomics
• Population Genetics
• Bioinformatics

Background:

• Public reference panels offer high imputation precision for European genomes.
• Study-specific panels (SSPs) can enhance imputation, but combining them with public panels is difficult using external servers.
• Geographic proximity between reference and target individuals is crucial for accurate imputation.

Purpose of the Study:

• To compare genome imputation accuracy using a public panel versus a French study-specific panel (SSP).
• To investigate scenarios favoring SSP-based versus server-based imputation.
• To propose a method for combining imputation strategies for improved accuracy.

Main Methods:

• Imputed 550 French individuals using the University of Michigan imputation server (Haplotype Reference Consortium panel) and an in-house French SSP (850 individuals).
• Analyzed imputation accuracy based on geographic proximity of target and SSP individuals.
• Investigated haplotype sharing and population fine-structure in France.

Main Results:

• SSP-based imputation showed benefits for haplotype phasing and rare variant imputation (MAF < 0.01), particularly for geographically matched individuals.
• Imputation accuracy improved to 58.1% for individuals from regions well-covered by the SSP, compared to 42.3% overall.
• A pragmatic approach combining server-based and SSP-based imputation results via posterior genotype probabilities achieved higher accuracy than either method alone.

Conclusions:

• Study-specific panels offer significant advantages for imputing French genomes, especially for rare variants and when geographically relevant.
• Combining imputation strategies is essential for maximizing accuracy, particularly when direct combination of panels is not feasible.
• The findings provide insights into expected imputation accuracy across different strategies for European populations.