I am doing a discriminative analysis using MEME tool in sequences that are changing DNA methylation. If I have a motif that is present with a probability of 60% in sequences of interest and 34% in control sequences. Could I say that this motif is related to DNA methylation changes? is there a threshold to make conclusions? for example a 25% increase in the presence of this motif means this motif is involved but a 20% means this is not so probable, etc. Any reference? Any recommendation about what things I should consider in the analysis would be really appreciated. Thanks
Hi!
You could do this in two ways. Either way, you will need to make sure your control sequences are robustly selected and appropriate for the question you are trying to ask. We can further discuss this if you want in the context of your experiment.
1) Testing for enrichment of known motifs
Use FIMO to search for each motif in both target and control sequences and then use something like Fisher's statistic to test for enrichment.
This approach is very sensitive to the p-value threshold used in FIMO. The default (1e-5) is too strict for my taste. Depending on how degenerate your motif is, something between 1e-3 and 1e-4 should be more appropriate.
2) De-novo motif discovery
Use meme-chip with both target and control sequence files. You should mind however that by default, meme-chip does not use all your sequences and in addition it slices a portion of each "peak" centered around the middle.
I would also use Homer alongside meme-chip and compare the results.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics