GWAS using 1kg or other database as "controls"?

Question

[Sean Davis]

We have a small study (100+ cases) in which we did not collect controls (main interest is in somatic variation). What approaches would folks suggest trying in such a situation to do a GWAS-lite analysis (this would be a hypothesis-generating exercise....)? In particular, can 1kg data be used for the controls?

These are naive questions, I know, but our study design leaves a bit to be desired and I haven't found a good reference that matches our situation.

[Katie D'Aco]

You say you're interested in somatic variants. Are you looking at paired tumor/normal data?

[Sean Davis]

We have paired tumor/normal data, yes. The GWAS is to look for disease predisposition in the germline genotypes.

Answer 1

[Katie D'Aco]

Using public databases as controls can be appropriate if your phenotype of interest is very rare or severe. Because of your small sample size, I would use the public data to filter out variants that occur in your data instead of running any statistical tests. You could also do some hypothesis generation by asking questions like "what genes have a somatic coding/non-synonymous/etc variant in at least two cases but in no controls."

You mentioned 1KG, but I would also check out the Exome Variant Server

Edit: I just saw your comment about looking for germline cancer risk variants, which I think will be difficult. Maybe start with a list of known cancer risk genes to see if there is anything obvious there.