Are there good benchmarking/validation tools for CNVs? By 'benchmarking/validation', I mean, you feed the tool a truth set and the CNVs called by a CNV caller, then the tool will output something like the sensitivity and specificity of the CNV caller.
Thanks!
•
written
by
Jackie
Hi, did you try Ximmer ?? It is "Ximmer is a tool designed to help users of exome and targeted genomic sequencing data accurately detect and interpret copy number variants (CNVs). Ximmer is not a copy number detection tool itself. Rather, it is a framework for running other copy number detection tools and interpreting their results"
Hi,
The framework CNVbenchmarkeR was implemented with this purpose. Currently, it supports DECoN, CoNVaDING, panelcn.MOPS, ExomeDepth and CODEX2 tools. Hope it helps.
We used it to benchmark CNV detection tools for targeted NGS data (paper)
We have released a 2nd version of the framework, CNVbenchamarker2. It supports 12 tools: GATK-gCNV, DECoN, CoNVaDING, panelcn.MOPS, ExomeDepth, CODEX2, ClinCNV, clearCNV, Atlas-CNV, Cobalt, CNVkit and VisCap.
Also we published the results comparing the tools in a new work.
I would say comparison with Genome-in-a-Bottle results for a well-known human sample would be the best benchmark.
Log in to answer this question.
It depends on the data you use. For WGS, check out the GiaB (https://www.nature.com/articles/s41587-020-0538-8.epdf) HG002 truth set. There are also tools: truvari, SVanalyzer that compare the vcfs and report Precision/Recall metrics.
For WES, panels, you can try ICR96 dataset (you will need to request): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5473400/
HTH