Precise estimation of in-depth relatedness in biobank-scale datasets using deepKin

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genetics

Background:

• Accurate estimation of genetic relatedness is crucial for large-scale genetic studies and biobanks.
• Traditional methods often rely on fixed thresholds, which may not be optimal for diverse datasets.
• Understanding relatedness is key to controlling for population structure and identifying familial relationships.

Purpose of the Study:

• To introduce deepKin, a novel method-of-moments framework for accurate relatedness estimation in biobank-scale genetic studies.
• To develop a method that accounts for sampling variance to enable robust statistical inference and classification of relatedness.
• To provide tools for determining data-specific significance thresholds and estimating statistical power for detecting distant relatives.

Main Methods:

• Developed deepKin, a method-of-moments framework incorporating sampling variance.
• Implemented data-specific significance threshold computation.
• Determined the minimum effective number of markers and estimated statistical power.
• Validated through simulations and application to UK Biobank data.

Main Results:

• deepKin accurately infers unrelated pairs and relatives by leveraging sampling variance.
• Analysis of the UK Biobank 3K Oxford subset showed increased power for distant relative detection with larger effective marker sets.
• In the UK Biobank White British subset, deepKin identified over 212,000 significant relative pairs across six degrees.
• Geographic patterns of relatedness were revealed across 19 UK Biobank assessment centers.

Conclusions:

• deepKin offers a statistically rigorous approach to relatedness estimation in large genetic datasets.
• The method enhances the power to detect distant relatives and provides insights into population structure and familial relationships.
• An R package (deepKin) is available for broader application in genetic research.