MEME suite compare motifs between species

Question

[Rafael Soler]

Hello,

I have a set of 5kb sequences upstream of a gene from different primates, and I would like to know what motifs are in enhancers and promoters of primates vs the 5kb upstream sequences of a set of the same gene but for rodents.

Would the analysis with MEME suite be correct with the "Discriminative mode", and putting the rodent sequences in control sequences and the primate sequences in the primary sequences? To contrast, since the normal method looks for similarities.

If not, how would this analysis be carried out?

Thanks!

Answer 1

[Alex Reynolds]

I think the discriminative mode is more to look for enrichment of the same motif. But different species may have the same motif with different information values.

So one thought is, given sets of orthologous genes, you might perhaps perform a MEME scan for conserved motifs over their respective upstream promoter sequences. Pairs of per-species MEME PWMs could be compared with TOMTOM to query for significantly similar motifs. Putting results into a grid could help with comparing all species pairings at a glance.

Another thought is to use existing (MEME-formatted) PWMs with FIMO to generate whole-genome scans for published motifs, comparing hits over ortholog promoter regions.

[Rafael Soler]

Hi there! Thanks for answering Alex,

The problem with doing normal MEME is that it finds the same motifs in rodent and primate sequences (which it gives as significant). However, if you look at them in sequence, you can see how they are similar but clearly they are two different sequences, which happens that in normal MEME the algorithm "struggles" to find similarities. I was looking for some kind of analysis to find the differences between them.

If I perform the two analyzes separately (primates and then rodents), they do give me different predicted motifs. Would it be better to analyze them separately and compare the PWMs?

All the best,

Rafa

[Alex Reynolds]

You might compare the per-species PWMs against each other, or each separately against published motif databases, using TOMTOM (part of the larger MEME toolkit). With the similarity scores you get back, you might see how quantitatively close motifs are to each other, or get a sense of similarity to motifs of known TFs, which may together be in a larger TF family or part of a regulatory network of interest.

[Rafael Soler]

Thank you so much! :)