CrowdVariant: a crowdsourcing approach to classify copy number variants.

Area of Science:

• Genomics
• Bioinformatics
• Human Genetics

Background:

• Copy number variants (CNVs) are crucial genetic variations linked to numerous diseases.
• Accurate identification of CNVs, especially small ones (<10kbp), remains challenging using array and next-generation sequencing (NGS) data.
• Expert manual curation is the gold standard but lacks scalability for clinical sequencing demands.

Purpose of the Study:

• To demonstrate the feasibility of high-throughput manual curation of CNVs by non-experts.
• To develop and validate a crowdsourcing framework (CrowdVariant) for CNV identification.
• To create a high-confidence set of deletions for a reference sample (NA24385/HG002).

Main Methods:

• Development of the CrowdVariant crowdsourcing framework utilizing Google's platform.
• Recruitment and training of non-expert annotators for CNV classification.
• Validation of non-expert classifications against expert calls and established datasets.

Main Results:

• Non-expert annotators demonstrated high agreement with each other and with expert classifications.
• Crowdsourced classifications accurately assigned copy number status to putative CNVs.
• Identified 1,781 high-confidence deletions in the NA24385 reference sample, surpassing existing callsets.

Conclusions:

• Crowdsourcing offers a scalable solution for manual CNV curation, addressing limitations of expert-only approaches.
• The CrowdVariant framework shows clinical potential for large-scale CNV analysis.
• This methodology can enhance CNV calling algorithms and guide future crowdsourcing efforts in genomics.