Entering edit mode
6.2 years ago
rbronste
▴
420
Hi so Im doing some motif searches and was hoping to get some clarification on how these programs work. When running AME on a set of 250 sequences I get a certain number of hits back and for one specific motif/PWM it was present in 16 of the fasta sequences with some of those being true positive hits and some being true false positive. However when running the same exact PWM against the same sequences through FIMO it found 51 occurrences of this motif, wondering how they differ in this sense? Thanks!
I do not know the algorithms in detail but here are a two things to consider:
Cutoffs: Typically, a p-value is calculated and used to select true from false/random positives. Check out if the model is the same (e.g. Binomial vs. Hypergeometric) and if there is a multiple-testing correction. FIMO definitely has one, I think they use an FDR-corrected p-value of 1e-4 as default. If so, which one: Might be Bonferroni or Benjamini-Hochberg. Also check if the default cutoffs are somewhat comparable.
Background model: As stated in the manual of FIMO and some posts in its google mailing list (I do not have the links ready now), it was clearly stated that the background model (background regions representing a similar nucleotide composition) has a notable impact on the output. Check which models the tools use and if they are comparable.