Construction of relatedness matrices using genotyping-by-sequencing data

Area of Science:

• Genomics
• Population Genetics

Background:

• Genotyping-by-sequencing (GBS) offers a cost-effective alternative to array-based genotyping for single nucleotide polymorphisms (SNPs).
• Reducing sequencing depth lowers costs but compromises genotype call quality, often necessitating stringent SNP filtering and reducing data utility.
• Existing methods struggle with low-depth GBS data, limiting its application in relatedness estimation.

Purpose of the Study:

• To develop and evaluate methods for accurate relatedness estimation using low-depth GBS data.
• To investigate the impact of sequencing depth on genotype calls and relatedness estimates.
• To provide a robust framework for analyzing GBS data in population genetics studies.

Main Methods:

• Theoretical calculations and simulations were employed to assess relatedness estimation methods.
• A novel approach, kinship using GBS with depth adjustment (KGD), was developed to account for varying SNP depths, including zero calls.
• Matrix methods were utilized for efficient computation of relatedness estimators.

Main Results:

• Unbiased relatedness estimates are achievable using SNPs genotyped in both individuals, independent of individual sequencing depth.
• A modified estimator provides unbiased self-relatedness estimates, accounting for SNP depth.
• Optimal sequencing depths were determined: 2-4x for individual relatedness and 5-10x for self-relatedness.
• A graphical method ('fin plot') aids in filtering SNPs exhibiting non-Mendelian behavior.

Conclusions:

• The KGD method enables accurate relatedness estimation from GBS data, even with low or zero depth genotype calls.
• This approach allows for optimized sequencing depth selection in GBS experiments.
• A simple graphical filtering method effectively identifies and excludes problematic SNPs, enhancing data quality.