SurVIndel2: improving copy number variant calling from next-generation sequencing using hidden split reads

Area of Science:

• Genomics
• Bioinformatics
• Structural Variation Analysis

Background:

• Copy number variations (CNVs), including deletions and tandem duplications, constitute the majority of structural variations in the human genome.
• Existing short-read sequencing methods struggle to detect CNVs in repetitive genomic regions due to insufficient evidence (split reads, discordant pairs, read depth changes).

Purpose of the Study:

• To introduce SurVIndel2, a novel short-read-based caller for detecting CNVs, particularly those in challenging repetitive regions.
• To demonstrate SurVIndel2's superior performance compared to existing callers on human and non-human datasets.
• To enhance the comprehensiveness of human genome variant catalogs by identifying previously missed CNVs.

Main Methods:

• Development of SurVIndel2, incorporating novel 'hidden split read' evidence alongside established statistical techniques.
• Benchmarking SurVIndel2 against popular CNV callers using public datasets.
• Application of SurVIndel2 to the 1000 Genomes Project dataset to generate an extensive CNV catalog.
• Integration of SurVIndel2 with Google DeepVariant for combined indel and CNV variant calling.

Main Results:

• SurVIndel2 significantly outperforms existing callers in detecting CNVs, especially in repetitive regions.
• A comprehensive CNV catalog for the 1000 Genomes Project was generated, revealing hundreds of thousands of previously undetected variants.
• Tandem use of SurVIndel2 and DeepVariant produced a remarkably complete catalog of variants for an individual genome.
• Limitations in current sequencing technologies were identified as a major contributor to missing CNVs.

Conclusions:

• SurVIndel2 represents a significant advancement in CNV detection using short-read data, particularly in complex genomic regions.
• The method substantially improves the completeness of structural variant catalogs, aiding in a deeper understanding of human genetic diversity.
• Addressing sequencing technology limitations is crucial for further improving variant detection accuracy and completeness.