Rapid genotype refinement for whole-genome sequencing data using multi-variate normal distributions

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• Whole-genome sequencing (WGS) combined with genotype refinement is crucial for large cohort studies.
• Existing genotype refinement methods, often based on hidden Markov models, are accurate but computationally intensive.

Purpose of the Study:

• To introduce a novel, computationally efficient algorithm for genotype refinement.
• To improve the speed and scalability of inferring genotypes from low-coverage WGS data.

Main Methods:

• Developed a new algorithm for genotype refinement that models linkage disequilibrium (LD) using a multivariate Gaussian distribution.
• Implemented a computationally efficient approach that avoids the complexity of hidden Markov models.

Main Results:

• The proposed algorithm is hundreds of times faster than existing methods.
• The method exhibits linear scaling with the number of samples, enhancing its applicability to large cohorts.
• Demonstrated performance on both low- and high-coverage sequencing data.

Conclusions:

• The novel Gaussian distribution-based algorithm offers a significant speed improvement for genotype refinement.
• This method provides a cost-effective and accurate solution for genotype inference in large-scale genomic studies.