Help With Fisher Test For A More Than Two Group Of Genes‏

Question

[rantree]

Good Afternoon,

My name is Gabriel, I'm doing an analysis if there is increase or decrease in dependence on the mutated genes, using 3 or more genes using the fisher exact test. I performed with success an analysis for two genes using fisher.test( ). example of the 2x2 contigency table:

                                    Gene A mutated | Gene A normal
Gene B mutated|              26                    |        12
--------------------------------------------------------------------------------------------
Gene B normal |             10                      |          50

Now I'm wondering how can I perform the analysis for 3 genes (and construct the contigency table), as follows: Gene A mutated, Gene A normal, Gene B mutated, Gene B normal, Gene C mutated and Gene C normal. How do I perform a fisher test using fisher.test( ) function using this data (3x3 contigency table)? Can someone help me ?

Answer 1

[jockbanan]

The problem here is that Fisher test is meant for categorical data. That means that each of your patients must belong to exactly one category. In your case, you have 3 variables, each with 2 categories. This implies a 2x2x2 "cube" matrix, so that e.g. in one cell, there will be the number of patients that have mutated gene A & normal gene B & mutated gene C.

A 3x2 table would be for problems like: man - woman - alien vs. 10 fingers - more fingers i.e. two variables, three and two categories, respectively.

Common fisher test is not meant for such problems (see nice short definition here: http://mathworld.wolfram.com/FishersExactTest.html) I'm not sure, if there is a program that could compute this 2x2x2 Fisher test modification, but googling gives me some results.

Answer 2

[Ali R. Vahdati]

If you have a 3x3 as a matrix, you just use fisher.test(yourTable).

[rantree]

How do I construct this 3x3 matrix? I have 3 genes and two states

Answer 3

[polarise]

I think you should use a 3x2 (2x3) contingency table because you have three genes and two states (mutated vs. normal).

[rantree]

but if do like this : Columns : Gene A, Gene B and Gene C , rows : mutated , normal . And if I assume that we are analyzing 3 patients.... one patients can have Gene A and Gene B mutated but not C.... another patient Only Gene C and the other patient all three genes mutated, at the end will have an contingency table with :

              gene A | gene B | gene C
mutated     2           2             2
normal       1            1             1

a total of 6 mutated and 3 normal....and our n of patients is three ....not 9!