I did a Chip Seq analysis and I used a MEME program to analyze the consensus motif for the transcription factor binding site. I obtained a long sequence (24 nucleotides). I did a search of this sequence and I found that it is one of the most frequent n-mers in the human genome.
What does this mean exactly? Is there any way to know if my consensus motif is real or only a statistically significant repetition?
MEME will find an overrepresented subsequence in a set of sequences. However that sequence could well be overrepresented in the genome as a whole, so that it will be foun overrepresented in any subset of sequences.. This MEME analysis should always be followed by a specificity analysis, i.e., are these motifs specific to MY subset of sequences? I think there are other tools that could help you with that as part of the MEME/MAST suite. Hope this helps.
MEME will find an overrepresented subsequence in a set of sequences. However that sequence could well be overrepresented in the genome as a whole, so that it will be foun overrepresented in any subset of sequences.. This MEME analysis should always be followed by a specificity analysis, i.e., are these motifs specific to MY subset of sequences? I think there are other tools that could help you with that as part of the MEME/MAST suite. Hope this helps.
I agree with Diego; the issue with MEME is that it's very good if you can access it through a cluster computer to override the website limitations and input as many sequences you want, but then, it makes sense only if the underlying biology does...
this is well known problem in motif detection. That's why most motif detection tools take a background model to compare to. Usually, the background model is drawn from the entire genome sequence, the UTRs, intergenic regions, etc. Most background models take into account the CG content of the background sequences, but some also take into account the frequency of double or triple letters (1st, 2nd, or 3rd order models). MEME gives you the option to either upload a background model or a negative set of sequences. You should take advantage of this.
If you did ChIP-seq against transcription factor antibody, you should get known motif for the transcription factor. First check your ChIP-seq peaks globally via genome browsers such as UCSC genome browser or IGV. If you see clear peaks over many regions, then extract flanking sequences (let say -50 bp ~ peak center ~ +50 bp) of top 1000 peaks (ranked by FDR or fold-change based on your peak calling program) and use it as MEME-ChIP input. You should get conventional DNA binding motif of the transcription factor. This strategy works well for transcription factor ChIP-seq. Also check how many sites contain the identified long sequence over your input. If the number is small, then it may be meaningless.
Hello,
I think that the better way to answer your question is to assay the DNA binding activity of your protein with the consensus DNA sequence obtained. If you've got the possibility to construct some site mutants of this 24 bp binding region would be even better. MEME can work well, however in my experience there is a
considerable background in some cases.
As mentioned, the problem is that using a control set of sequences is difficult with Meme. There are two types of motif analysis: de novo creation and enrichment calculation. Might I suggest something like TRAP multi-sequence (http://trap.molgen.mpg.de/cgi-bin/home.cgi) which does enrichment calculation over a control. RSAT's (http://rsat.ulb.ac.be/) peak-motifs tool does a de novo motif creation with background and then compares the resulting PWM to a library of known motifs
Thank you all for your time! I think I'm going to continue with an EMSA to assay the binding activity of my protein and this DNA sequence. As Marcelo advised me, I'll use a mutant in the sequence as control. Do you think is a good idea?
- Badges
- Science topic
- Similar topics
- Biological Science
- Genetics