Optimized sample selection for cost-efficient long-read population sequencing

Area of Science:

• Population Genetics
• Genomics
• Bioinformatics

Background:

• Large cohorts are often genotyped with low-resolution methods, limiting captured genetic diversity.
• Resequencing a small subset with high-resolution methods (e.g., long-read sequencing) requires careful sample selection for representativeness.
• Historical genetic studies often overrepresent specific ancestries, neglecting global diversity.

Purpose of the Study:

• To develop a computational tool, SVCollector, for identifying an optimal subset of individuals for deep resequencing.
• To maximize the representation of genetic diversity and variants across all subpopulations within a selected subset.
• To address biases in genetic diversity sampling in large cohort studies.

Main Methods:

• SVCollector analyzes population-level VCF files from low-resolution genotyping data.
• It employs a greedy heuristic and an exact integer linear programming algorithm to solve the subset optimization problem.
• The tool ranks samples to maximize variant discovery within a specified subset size.

Main Results:

• SVCollector identified more representative sample subsets compared to naive strategies in simulated data, the 1000 Genomes Project, and the 3000 Rice Genomes Project.
• When selecting 100 samples, SVCollector included individuals from every subpopulation, unlike unbalanced naive methods.
• The number of variants discovered in SVCollector-selected cohorts follows a power-law distribution related to the allele frequency spectrum.

Conclusions:

• SVCollector provides an effective method for optimizing sample selection in population genetics studies.
• This approach enhances the capture of overall genetic diversity and improves variant discovery.
• The method offers a way to estimate population diversity more accurately with increasing sample sizes.