Fisher's exact test

• For experiments without replication
• Fisher's exact test can be used on a 2x2 (or larger) contingency matrix to test dependency between rows and columns
• For expression counts, the 4 values are
  • number of counts for Gene A in Sample 1
  • number of counts not for Gene A in Sample 1
  • number of counts for Gene A in Sample 2
  • number of counts not for Gene A in Sample 2
• A test is performed for every gene of interest.
• This was a popular method in earlier SAGE expression counts and is also tested in ​Bullard et al., 2010.
• We don't yet have a command-line script to perform this on 2 samples, followed by FDR correction.
• As with any method not taking advantage of biological replication, any conclusions must be treated with care (and not taken too seriously).