deepKin

deepKin estimates genetic relatedness using a method-of-moments framework that accounts for sampling variance to enable statistical inference and classification of relatedness in biobank-scale genetic studies.

Key Features:

• Data-specific significance thresholds: deepKin computes significance thresholds tailored to each dataset rather than using fixed thresholds.
• Minimum effective marker determination: The framework determines the minimum effective number of markers needed for analysis.
• Statistical power estimation: deepKin estimates the statistical power to detect distant relatives.
• Relatedness classification: The method classifies relative pairs into degrees of relatedness, including identification of unrelated pairs.
• Within-cohort and cross-cohort analysis: The framework supports relatedness estimation both within and across cohorts.

Scientific Applications:

• Biobank-scale genetic studies: deepKin improves accuracy of relatedness estimation in large-scale genetic studies such as the UK Biobank.
• Distant relative detection: The method detects distant relatives by leveraging SNP sets with larger effective numbers of markers to increase power.
• Population structure analysis: deepKin reveals geographic patterns of relatedness across assessment centers, as demonstrated with 19 UK Biobank assessment centers in the White British subset.
• Relatedness degree categorization: The framework categorizes significant relative pairs into six degrees of relatedness, as shown by identification of over 212,000 significant relative pairs in the White British subset.

Methodology:

deepKin uses a method-of-moments framework that accounts for sampling variance. The method computes data-specific significance thresholds, determines the minimum effective number of markers, and estimates statistical power for detecting distant relatives.