The
theory of this association test may be found in the manuscript listed
below (Zhang et al 2008). The essence of this approach is a feature
analogous to Holman's triangle. Consider a 2 x 3 table with rows
corresponding to cases and controls, and columns representing the
three genotypes of a SNP, AA, AB, and BB. The table body contains
numbers of individuals. A chi-square association test of these data
has 2 df, which may be represented by q1 =
frequency of AA
genotypes, and q3 = frequency of BB genotypes.
For the FP
test, the 2 df are re-formulated in terms of SNP allele frequency p
and inbreeding coefficient F. The equation F
= 0 is a
curved line in the plane spanned by q1 and q3,
and biologically reasonable values, F > 0,
correspond to
2/3 of the unrestricted parameter space (q1, q3).
It is this restriction that makes the FP test more powerful than the
2 df chi-square test, particularly under recessive inheritance and
when cases show more homozygotes than expected under HWE.Download the program package and move its contents into a folder. Preferably, run the FPtest program in a Windows command window (CMD), or double-click on FPtest (FPtest.exe).
The program will ask you to furnish, for a given SNP, numbers of individuals with genotypes AA, AB, and BB, both for cases and for controls. It will then compute p-values via permutation samples, with the observed data being counted as a null dataset. With n = 1,000,000 permutations (default), the smallest p-value that can be obtained is 1/(n + 1) = 10-6, where n is the number of permutations performed. The program will write output to a file called FPtestout.txt.
Permutation analysis will require a seed for the random number generator. The program will create this seed based on the system clock unless you provide a file called seed.txt containing a positive integer number. The FPtest has also been implemented in our sumstat program package.