How to determine if SNP is associated with disease - (Jul/14/2013 )

Let's say you sequence a certain gene that is associated with a disease. How do you determine whether the specific allele that you sequenced is one of those disease-causing polymorphisms?

I guess you would have to find incidences where it is the only mutation (or one of a group of) and find that in a number of specimens of the disease. Epidemiology I guess.

You can't tell it for sure unless you have a proof. Usually overrepresentation in diseased population is not sufficient, but there must be functional studies done on this specific variant.

First you can rule out if it's not a disease causing, because that is more simple. If it's a silent mutation, change that it's a functional SNP is quite low (though can't be ruled out, but you'd need to press much more evidence, requirement for regulatory functions or splicing deviation). Then you check dbSNP if it's an already registered SNP, and if so, due to the 100 genomes project, it would have also population frequency. The higher the frequency the lesser the change it has a negative function.

You can then look closer into the protein region containing SNP, what function that domain has, were any published causative mutation also in this domain. You can use protein prediction applications like SIFT or Polyphen to determine if it's likely this change would have any effect on the protein (SIFT does alignment to all known homologous proteins in all available organisms, and determine whether this residue is conserved and thus expected to be important, Polyphen analyses AFAIK the properties of the AA change and computes some possible changes in protein folding).

If you assume from the computer analysis, that it's an unknown variation in a region that may play role and predictions ale mostly in favour of negative effect on protein, it may be worth to spend additional money on functional experiments.