Minimum number of sequences for SEA enrichment analysis

[Manuel Fernández Merino]

Hello,

I am trying to look for enrichment on sequences on a very specific pattern observed in a Hi-C experiment. That means I'm using 1kb-long input sequences (my Hi-C experiment resolution), and that I'm checking less than 10 of such sequences. Would such analysis be reliable? I am worried that the low number of sequences could make this analysis pointless.

Thanks,

Manuel

[cegrant]

I believe that it should be possible to get meaningful
results when analyzing 15 1Kb sequences (or more) with SEA.  The level
of statistical significance would depend partly on the number of multiple
tests you are doing.  Effectively you are doing one test for each motif
in your input.  It also will depend on whether a sizeable fraction of your
sequences have a strong match to the motif.

My best advice is to try it, and see what kind of Q-values you see.

(Tim Bailey)