Can anyone suggest a method to calculate a sample size to study an effect of epistasis between two polymorphisms over a clinical phenotype (ex: drug response). Suggestion of any reference will be appreciated.
To calculate sample size and power estimates are the same with genetic data as with other traditional data types, but allele frequencies need to be taken into consideration when using genetic data. Typical sample size calculations yield a number of subjects, usually per group, that are necessary to detect a significant difference between those groups. If a calculation indicates a sample size of 50 per genotype group is required, the allele frequency of the minor allele then needs to be considered to calculate how many total subjects you need to get 50 heterozygous mutants (the rarest genotype).
For example, a sample size calculation might determine that you need 50 homozygous wild-type individuals (AA) and 50 homozygous mutant individuals (BB) to adequately test a hypothesis.
You might want to have a look at Shuang Wang & Hongyu Zhao (2003): Sample Size Needed to Detect Gene-Gene Interactions using Association Designs. Am. J. Epidemiol. 158 (9): 899-914. doi: 10.1093/aje/kwg233
And maybe you'd also like to try out Quanto (http://biostats.usc.edu/Quanto.html), a power analysis program for Genetic studies, which also provides a mode for GxG interaction analysis