How to conduct pathway analysis from VCF file?

Question

[David_emir]

Hello All,

I have performed variant calling on re-calibrated bam of my Exome seq samples,

java -jar GenomeAnalysisTK.jar -T HaplotypeCaller -R ref -I recal_reads.bam -o raw_variants_recal.vcf

Now my question is - After annotating using SnpEff, can i submit those genes to get pathway analysis done? of yes how to do that?

Thanks a lot Dav.

Answer 1

[Manuel Landesfeind]

Yes, you can do pathway analyses - the issue is that you should first think about what your research goal is and how any kind of pathway analysis can contribute!

For example, you can use a list of genes that have a mutation and you can use this list to do an over-representation analysis on pathways (a.k.a. Fisher's exact test). The results will identify pathways with an unexpected high number of mutated genes.

Also, you can annotate your SNP with an impact score (must be numerical! SnpEFF is not sufficient) and can do a Set Enrichment Analysis or Functional-Class Scoring using pathways as sets.

Third, if you have very well annotated SNPs and using a sophisticated database you can also do impact analyses based on pathway topology. For example, knowing that a mutation results in a protein activation (like BRAF V600E) you can infer that the pathway downstream should show a strong signal. This could lead to the conclusion that you would expect an increase in transcript of the target genes of this signaling pathways. However, for such methods you would need more knowledge on the mutations as well as on the exact signaling cascades of the related pathways that is usually not available in public databases (at least not easily accessible)...

However, I strongly recommend to discuss your plans with someone who has some experience in this types of analysis. Given your very broad question on "how can I do pathway analysis", I assume that you lack any experience so far and I strongly recommend to study some literature to first learn

• what kinds of pathway analysis are possible
• what questions they answer
• what are the benefits and drawbacks of the individual methods

Afterwards you can think about IF any type of pathway analysis will give you any result that is related to your research after all...

Answer 2

[Collin]

You can submit variants in VCF format to CRAVAT (http://www.cravat.us/ ) and it can score variants with VEST and do a pathway enrichment of high scoring genes (if you examine the interactive variant viewer).

[David_emir]

Thanks Collin, can i have .vcf files from Rat genome, can i use this in CRAVAT?

[Collin]

Unfortunately, CRAVAT is human only.

Answer 3

[willgilks]

Hi David,

Your question seems a bit strange. You have genotype data, and now you want to find out which proteins interact .. and you want to apply a pathway analysis but don't know what to apply it on.

Perhaps, the missing step in your plan is to calculate which genetic variants are correlated with a phenotype. Then you can perform a pathway analysis on the significant variants. Basically you need some way of catagorising your genes so you can do a pathway analysis.

One interesting approach might be to look for long-range linkage-disequilibrium, including between chromosomes. This could identify some evidence of protein-protein interactions but I don't know if this has been developed beyond theory.

In general, I would ask your supervisor for guidance as your question still seems strange.

[David_emir]

thanks Willgilks for your insights , what i mean is not Protein Protein interactions exactly, can i do a pathway analysis on the annotated data, just to know the genes involved in the pathway. i just got this idea from RNA seq experiment where in we can do a pathway analysis,just wondering if we can do some thing like that from Exome seq data... I may be wrong, but wanted to explore more on this.

Answer 4

[willgilks]

RNA seq quantifies the expression level of each individual gene. Exome sequencing does not. Therefore it is not possible to do pathway analysis of expression levels from exome sequencing data.