Exome sequencing and association

Question

[sailepradh]

Hello all,

I have some fundamental question regarding the association of a phenotype (drug toxicity) with the genetic variants. For the association of the phenotype, we had done exome sequencing of 200 patients and got genetic variants of individual patients after processing through the GATK pipeline. Are there any statistical tools that can help in the association studies of the both common and rare variants founds with the phenotype? Since we are considering both rare and common variants in association of the disease and also since the sample size is only 200, the normal single variant test lead to low association power. Here the phenotype in consideration are the haematological toxicity with the quantitative values.

[ivivek_ngs]

Your question is not particularly clear but what I can understand is from the SNP and INDELs you would like to know which are common and rare. You can check your variants against the dbSNP database and discount the ones already found there and for population you can look for EU, African and American population studies. Also take a look at the SNPsnap

[sailepradh]

Hi there,

I want to rephrase my question. Is there any statistical method that can associate the rare and common variants from the exome sequencing with the phenotype i.e toxicity in my case. Through the software I can differentiate between the common and rare variants and in fact I have done it but since there is lack of control I am looking for appropriate methods.

Thank you in advance

[Biomonika (Noolean)]

You should edit your original question rather than writing the edit as your answer. Now it might seem like your question already has one answer, although not accepted.

Answer 1

[raunakms]

I think your question is still very unclear! What are the phenotypes your are considering? Are you phenotypes qualitative or quantitative?

• If it's qualitative (like group comparison) you would at least need two groups to compare. A simple chi-square test would be good enough to test if the variant is significant to the phenotype group or you could even build a simple classifier based on you data.
• However if the phenotypes are quantitative (like drug sensitivity) then you would have to build a regression based classifier or any other classifier based on model appropriate for your data. Since you mentioned you do not have control group, you could perform cross-validation on your data for validation purposes.

Refer to:

• http://www.nature.com/ng/journal/v46/n8/abs/ng.3021.html
• http://bioinformatics.oxfordjournals.org/content/25/12/i21.short

Answer 2

[muzilijiao]

I think you can perform SKAT anlysis to detect the association of a phenotype (drug toxicity) with the common and rare genetic variants.