What Is Your Experience With Promoter Prediction Programs?
3
5
Entering edit mode
13.3 years ago

Just a very quick question on promoter prediction software. Has anyone used any of the available tools and if so do you have any recommendations on a) the fastest b) the most accurate, and c) the best of both?

Here is a list of some I found! Are any of these scalable to whole-genome data? Are there any others not included here?

Of course if promoter prediction programs aren't the best way to retrieve these data (i.e. promoter sequences +/- first exon), then I'm open to suggestions :) I'm looking to do this for Hymenoptera (i.e. ant, bees, wasps).

Cheers,

Steve

promoter genome prediction gene software • 7.6k views
ADD COMMENT
2
Entering edit mode

Check out recent work by Saurabh Sinha at UIUC and maybe contact him directly. He works on predicting regulation in Fruit Fly and also Honey Bee. I think that he would have some insight on how to do it in those species.

ADD REPLY
1
Entering edit mode

I assume you mean human (or at least mammalian) but it would be helpful to specify.

ADD REPLY
0
Entering edit mode

My apologies, like I said a quick question :S I'll edit to clarify :D

ADD REPLY
0
Entering edit mode

For your information I'm looking at doing this for Hymenoptera and possibly Isoptera?

ADD REPLY
1
Entering edit mode
13.3 years ago
Vitis ★ 2.5k

I've used an approach called phylogenetic printing years ago and I'm sure it has new updates now. Basically, this algorithm takes in homologous non-coding sequences (for example, 5'-flanking regions of the same gene from multiple species) and tries to find evolutionary conserved DNA elements (short stretches of nucleotides) based on sequence conservation and species phylogenetic tree. In theory, these elements are potential regulatory elements but you never know until you actually do the transgenic experiments. With this approach, I managed to find TATA-box in a lot of cases, so it seems to be working.

ADD COMMENT
1
Entering edit mode

I used FootPrinter before, but it seems to be obsolete, only compiling on older Linux distros. I never tried BigFoot but it looks promising. So give them a try!

ADD REPLY
0
Entering edit mode

Thanks for this vitis :) I found a program called FootPrinter and a more recent one called BigFoot! Do you have any experience with these, or did you implement the algorithm into your own pipeline?

ADD REPLY
0
Entering edit mode

I'll have to have a play and see which ones work best - BigFoot does sound good though! I think by the sound of things though, any of these programs are unlikely to produce amazingly accurate results, in the absence of experimental evidence!

ADD REPLY
1
Entering edit mode
13.3 years ago

I have no experience with promoter discovery but an alternative approach would be to look for motifs directly upstream (and perhaps also downstream) of genes. I did a survey project on that for a Bioinformatics class.

You would need to have the genes discovered and moreover you would need to select sets of genes that may be co-regulated.

Weeder is one of the best motif discovery tools. Very fast and one of the beast in terms of accuracy (I can find the manuscript that did and exensive comparison of motif discovery tools and found Weeded to be the best). You have to give it a set sequences upstream of the genes that you think may have similar regulatory elements.

From what I've read, all unsupervised motif discovery methods still have very disappointing sensitivity.

On the Weeder web ui there is also a suggestion of Pscan (2009):

NEW If you are looking for over-represented motifs in promoter sequences, perhaps you can also find our brand new tool, Pscan useful.

ADD COMMENT
2
Entering edit mode

Aleksandr & Steve - the paper you are looking for is here: http://www.ncbi.nlm.nih.gov/pubmed/15637633. This paper does show that weeder is the best for motif discovery, but this is a biased conclusions since weeder is heavily trained on the transfac db, which is also used for evaluation

ADD REPLY
1
Entering edit mode

@gawbul, I searched for more than an hour (!) but I cant find this paper. It was an extensive survey - they looked at about 20 tools (MEME, MEME3, Weeder, consensus, ...) and found an expert for each one. They designed benchmarks and held the contest. Weeder tuned out on top. I think it was a 2005 paper. The fact that I can't find it means the paper was either really bad or something strange is going on.

ADD REPLY
1
Entering edit mode

@Casey. That's the one. Assessing computational tools for the discovery of transcription factor binding sites. Thank you! I didn't realize that Weeder was trained on transfac. Is that really true? In the survey/competition "consultation of TRANSFAC" was not permitted.

ADD REPLY
0
Entering edit mode

Hi Aleksandr :) Thanks for this, excellent answer (+1)! If you could find that manuscript it would be great, any idea what it was called or what journal?

ADD REPLY
0
Entering edit mode
ADD REPLY

Login before adding your answer.

Traffic: 1400 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6