Fisher exact test behavior when query tracks are very different

[ocha...@eng.ucsd.edu]

Hi all (but especially authors),

Figure 1d-e of the Giggle paper shows that Fisher's exact test is fairly concordant with MC estimation under some conditions; but I wonder

1) if anyone has done experiments varying the length and number of the intervals in one of the tracks relative to the other, and whether the approximation still holds; and

2) specifically, whether this estimate holds for a track of many (10k) small (10bp) intervals (TF motifs) against a track of fewer larger intervals, which is what I'm doing.

[Ryan Layer]

Great question. I think the best way to use GIGGLE is to identify a few tracks out of many that appear to be related to your query track, then use MC and many of the other stats tools to confirm.

--
You received this message because you are subscribed to the Google Groups "giggle" group.
To unsubscribe from this group and stop receiving emails from it, send an email to giggle-discus...@googlegroups.com.
To post to this group, send email to giggle-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/giggle-discuss/e42a4f65-d69f-49ad-ab79-7cd9de42a8db%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[ocha...@eng.ucsd.edu]

Using Genomic Hyperbrowser as a Monte-carlo based alternative, ran a query track against a 50 small TF motif binding site tracks (MC iterations at least 50).  Odds ratios are extremely concordant (R^2=.96, slope ~ 1).  P-values are concordant, but problematic if we do multiple test correction; when p-values are really small, we would need a lot of MC trials to precisely estimate p, which is prohibitively time-expensive.