Sequence Logos Comparisons

Hi! I have 3 groups of 15 sequence logos as shown in this image

http://tinypic.com/view.php?pic=n5iseb&s=6

Could I formally compare them somehow? For example the group in the right shows more conservative C positions than those in the left group and also the group in the middle can be distinguished for showing A's at low positions of the 16,17,18,19 width motifs..

But how could I formally show that differences or measure them?

Also if I could show the relations of the sequence logos like in here

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2425502/figure/F4/

but the method shown in the paper is too evolved for me.. is anybody aware of somehow to build similar trees?

Thanks

Ana

deleted-post

• 346 views • link

updated 13.7 years ago by Alex Paciorkowski

• written 13.7 years ago by anahochmanova •

1 answer

Welcome to Biostar.

I suggest you start with some reading on methods in multiple sequence alignment, particularly regarding identification of transcription factor binding sites. Are your sequences from different species? There are a number of tools out there that can manage what I think you are looking for -- statistically comparing conserved nucleotides among sequences.

The paper you reference uses a regression tree technique (called CSI-Tree) to identify previously uncharacterized transcription factor binding sites from what looks like microarray data. Are your data derived similarly? That paper's methods are available as an R package (here). If the method is too involved for you, perhaps you could contact the authors and discuss their technique? Perhaps they'd be willing to collaborate with you?

Overall, it sounds like you need a good bioinformatics collaboration to help you explore your data. Good luck!

• 0 views • link

written 13.7 years ago by Alex Paciorkowski

Thanks Alex, a bioinformatics collaboration is quite difficult because I don't have bioinformatics background, I'm just applying the techniques. I have identified different lengths motifs with the MEME algorithm and then represented them as sequence logos. Each of the identified motif is from one species that I have identified, I can't find how to select those motifs that somehow could predict the species.. Get some result like: motif TTTTTAT and TTTTTAA in 90% correspond to species Z. Could I identify relations like that?

• 0 views • link

written 13.7 years ago by anahochmanova •