Thank you to those who attended the recent webcast, “High Precision Exome CNV Detection with VS-CNV”. For those who could not attend but wish to watch, here is a link to the recording.
This webcast delved into the complex world of CNV calling for whole-exome samples, which presents unique challenges that require specific considerations and strategies. Over the past several months, our team has made numerous improvements to our CNV calling capabilities on exome sequencing data resulting in significant improvements to our algorithm’s precision. This webcast discusses these improvements and provides guidance on developing best-practice workflows for CNV calling from whole-exome coverage data. The improvements covered in this webcast include:
- GC correction and filtering
- Updated quality flags for filtering false positives
- Target quality assessment and filtering capabilities
- New settings for adjusting sensitivity and precision
During this webcast, I also compare the performance of our improved algorithm to that of the previous version in the context of two robust benchmark datasets to demonstrate the scale of the improvements to algorithm precision.
As with all of our webcasts, the final step was to answer viewer questions. Here are the responses:
Questions & Answers
Q: When looking at your benchmarking results, how did you determine the events that were false positive?
A: We utilized the Comprehensive Tier 1 Benchmark data published by NIST for assessing the number of false positives called by our algorithm. We considered any call made by VS-CNV that was within the Tier 1 Regions but not overlapping a call in the comprehensive Tier 1 benchmark set to be a false positive. The full text of the benchmark publication can be found here, and the data can be obtained from NCBI.
Q: When you have high coverage whole-genome sequencing data (~40X), would you recommend running the targeted or binned region approach to call CNVs?
A: The binned CNV caller is optimized to detect large events spanning multiple genes and is not capable of detecting small intragenic events. As such, for high coverage whole-genome data, we often recommend using the targeted CNV caller to ensure that small CNVs are not missed.
Q: Can you import externally called CNVs into VarSeq?
A: VarSeq can import externally called CNVs from the most common CNV calling technologies.
Q: Are there are any studies comparing this CNV detection method to other methods like MLPA?
A: Yes. In one such study, our clients at Robarts Research Institute analyzed 388 samples of patients with familial hypercholesterolemia, a disease caused predominantly by autosomal codominant mutations in the LDL receptor gene (LDRL). Their study was interested in seeing if the MLPA portion of the test could be replaced by our NGS-based CNV calling method. The authors found that the events detected by VS-CNV were in 100% concordance with those detected by MLPA. The full text of this study can be found here.
Q: Do you have CNV annotations like DGV?
A: VarSeq offers a comprehensive library of useful CNV annotations, including sources like DGV, 1kG CNVs, and Large Variants, GnomAD CNVs, ClinVar CNVs, ClinGen Dosage Sensitivity, and more.
Q: Thank you for the nice presentation! How much time does a CNV analysis take from BAM to called CNVs (with a medium-sized reference sample set)?
A: For a single exome sequencing sample with around 200,000 targets, Coverage Analysis and CNV calling typically take between 15 and 20 minutes.
The general purpose of any webcast is to showcase the capabilities of our software. Thank you for joining us for “High Precision Exome CNV Detection with VS-CNV” If you would like to have a more in-depth introduction to all GoldenHelix solutions, please contact us at [email protected]. Thank you again to all of those who attended, and we look forward to having you join us for our future demonstrations.
Feel free to check out some of our other blogs containing important, useful news and updates for the next-gen sequencing community.