Best way for CNV phenotype association analysis.

Question

[keremwai]

Hi, What would be the best way to do CNV (Copy number variation) association with a phenotype? Something like DESeq2/edgR for expression, or association of SNVs to phenotypes but for CNV. As a data source I have the Broad's CCLE (cancer cell line encyclopedia) and I want to associate CNV genetics with the reaction of the cell lines to a specific treatment. As a data source I either have for each cell line in the CCLE list of segments that are altered in their copy number, or I have the Broad's depMap data that includes a normalized to one (2 copies of the genome is one) copy number for each gene per each cell line. Is there a recommended package for this task? Or should I just do correlations or linear regression using R. Thanks Kerem

Answer 1

[svlachavas]

Dear Keremwai,

based on your description and project, I would suggest the R package CNVRanger, which provides various functionalities for CNV analysis, including a CNV-phenotype association pipeline, with a very comprehensive vignette:

https://doi.org/10.1093/bioinformatics/btz632

https://bioconductor.org/packages/release/bioc/html/CNVRanger.html

Best,

Efstathios-Iason

[keremwai]

Thank you Efstathios-Iason, I will try the package

Answer 2

[Garan]

I think if it's a case-control model then it would be a logistic regression? In this paper on arthrometric traits they use linear regression on CNVs (these would be mainly germline CNVs predicted from microarray data) and linear anthropometric traits such as height and weight in the UK Biobank dataset (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5622064/).

[keremwai]

Thank you. Would it be different for cancer somatic mutations?