0
$\begingroup$

I have a list of mutated genes from a VCF. Does it make sense to perform a pathway enrichment analysis using PANTHER or STRING as examples to know which pathway is affected?

EDIT: to reply to @terdon

We sequence mostly human genome data using Illumina sequencers. As for the type of data, we perform mostly Whole-Exome Sequencing (WES), Whole-Genome Sequencing (WGS).

At the end, after performing all pre-processing, we obtain the VCF files from HaplotypeCaller using GATK. Then we proceed to annotate using VEP.

Improve this question
$\endgroup$
6
  • $\begingroup$ That will depend entirely on what your data are. Did you sequence an entire genome? An entire exome? A small panel of target genes? Have you separated pathogenic from non pathogenic variants? How about compound heterozygotes? Please edit your question and give us some context to understand what you are trying to do. "A list of mutated genes" doesn't really tell us anything. $\endgroup$
    – terdon
    Jun 29 at 16:13
  • $\begingroup$ @terdon How and why does it matter what type of data is it coming from? Whether it's from an entire genome or exome panel? I can understand the question separating the pathogenic from non-pathogenic, but a mutation is a mutation whether coming into exons or introns, right? By compound heterozygotes, do you mean a multi-allelic mutation of a variant? $\endgroup$
    – user324810
    Jun 29 at 16:21
  • $\begingroup$ It matters because you want to do an enrichment analysis. If you're just looking at a small panel of genes, that would be a completely different proposition to looking at WGS or even WES data. If you are working with WGS data and have a set of affected samples and a set of controls, then enrichment analysis could make sense (but that would again depend on what you are trying to test for). However, you should first identify the variants (or genes with variants) that segregate with your affected samples. And by compound het I mean cases where you have two het variants in the same gene [cont...] $\endgroup$
    – terdon
    Jun 29 at 16:31
  • $\begingroup$ ... and it is the combination of the two variants that results in the affected phenotype, while neither of the variants would be sufficient by itself. $\endgroup$
    – terdon
    Jun 29 at 16:31
  • 1
    $\begingroup$ Yes, fair enough. I guess you can skip controls and assume simple over-representation. Just remember that in a WGS sample you expect ~5 million variants (including intronic and intergenic). Personally, I would use a collection of known healthy samples as controls, but you don't have to, you're right. $\endgroup$
    – terdon
    Jun 29 at 17:06
1
$\begingroup$

The concept of looking for enriched pathways makes sense, but you can't do it using the same tools you would use for gene expression data because the biases are different - longer genes are far more likely to contain variants by chance than short ones, and this needs to be accounted for in any enrichment analysis.

Improve this answer
$\endgroup$
2
  • $\begingroup$ But a mutation can also be silent on the gene. It should also be taken into account, right? $\endgroup$
    – user324810
    Jun 29 at 16:06
  • $\begingroup$ Yes, although you'd probably want to take that into account as well. $\endgroup$
    – Ian Sudbery
    Jun 29 at 18:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.