Population-scale detection of non-reference sequence variants using colored de Bruijn graphs.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput sequencing enables structural variant (SV) detection in numerous genomes.
• Non-reference sequence (NRS) variants are computationally challenging to detect using short-read data, often requiring de novo assembly.
• Existing algorithms for combining multi-genome data for NRS variant detection lack scalability for large datasets.

Purpose of the Study:

• To introduce PopIns2, a scalable tool for discovering and characterizing NRS variants in large cohorts.
• To present a novel algorithmic approach for merging de novo assemblies from multiple genomes.
• To demonstrate the performance and scalability of PopIns2 on simulated and real-world population-scale genomic data.

Main Methods:

• Developed a novel algorithm using a colored de Bruijn graph to merge contig assemblies of unaligned reads from multiple genomes.
• Integrated this merging approach into the PopIns2 tool for NRS variant detection.
• Evaluated PopIns2 on simulated data, the Polaris Diversity Cohort, and 1000 Icelandic human genomes.

Main Results:

• PopIns2 demonstrates significantly improved scalability for processing large numbers of genomes compared to previous methods.
• The novel merging algorithm achieves high quality and reliability in identifying NRS variants.
• PopIns2 exhibits superior precision as the number of processed genomes increases, outperforming existing approaches.

Conclusions:

• PopIns2 provides a scalable and precise solution for NRS variant detection in population-scale genomics.
• The tool's ability to handle large datasets opens new avenues for structural variant discovery.
• PopIns2 represents a significant advancement in the computational analysis of non-reference sequences across diverse genomes.