Comprehensive characterization of copy number variation (CNV) called from array, long- and short-read data

Area of Science:

• Genomics
• Bioinformatics

Background:

• Genome-wide high-throughput technologies like SNP arrays, short-read, and long-read sequencing are used to assay copy number variants (CNVs).
• Each technology has limitations and biases, some of which require further quantification.

Purpose of the Study:

• To compare the CNV calling capabilities of array, short-read, and long-read technologies.
• To validate CNV calls using raw data from each technology, rather than relying on a golden standard.

Main Methods:

• Assembled public datasets of CNV calls and raw data for the Genome in a Bottle individual NA12878.
• Included data from various methods and pipelines for CNV calling across array, short-read, and long-read technologies.
• Performed cross-technology comparisons and validated CNV calls using technology-specific raw data.

Main Results:

• Long-read platforms successfully identified CNVs in genomic regions not covered by arrays or short reads.
• CNV reproducibility across different pipelines within a technology correlated with CNV evidence measures.
• Distinct public database frequency profiles were observed for each technology, influenced by the database's underlying data source.

Conclusions:

• Long-read sequencing expands the scope of CNV detection into previously inaccessible genomic areas.
• CNV call reproducibility is a critical factor influenced by pipeline choice and evidence measures.
• The choice of technology significantly impacts CNV frequency profiles in public databases.