Identified Potential Non-Coding Rna, And Then?

Question

[Michael]

Assume we identify - by RNA-seq, tiling arrays, by prediction - possible candidate regions for non-coding, small RNAs. I wish to verify and predict the function of as many RNAs as possible by computation before going to the lab. One could use eg. Rfam to find similar sequences, after that we are left with more than 90% that have no match. One could predict the 2D structure using eg. RNAfold, compare that using RNAforester. But that does not get me even close to a function prediction. Do you have experience with other tools or a better computational pipeline that gets more information out of the ncRNA candidates, possibly even something specific to bacteria.

[Bioinfosm]

could you share the identifying part for ncRNA using RNA-Seq?

[Michael]

Just searched regions with high coverage not overlapping with coding sequences. As here

[Neilfws]

And then? Experimental verification :-)

[Michael]

@neil: so true ;) and of course, you cannot discover something that's not already annotated by a db search. unfortunately, my pipetting skills are rather underdeveloped.

Answer 1

[Fred Fleche]

[?]

http://www.bioinformatics.fr/resources.php?tag=non-coding-rnas

[Brad Chapman]

Great resource. After the site redesign, it looks like the link has changed to this.

[Jaydon]

This is a good approach to start with. We did similar thing before heading to wet lab.

[brentp]

that's a good resource i hadn't heard of!

[Fred Fleche]

The design is ugly but I am working on it (should be available in one or two months). Most of the ressources are from the NAR issues related to databases and webservers but organized by topics.

[Michael]

Great resource, thank you. I will try with my data.

[Fred Fleche]

Thanks for the remark I did update my response with the new link

Answer 2

[Paul Gardner]

How are you doing your comparison to Rfam? If you're just running rfam_scan.pl or the CMs then this may not give you the results you really want. See the recent paper by Kolbe & Eddy to hear more about the limitations of CMs on truncated sequences. Could explain some of your lack of sensitivity vs Rfam. However there is a terrifying number of ncRNAs not yet covered by Rfam.

I'm not sure clustering RNAfold predictions with RNAforester will tell you much either. Locarna and CMfinder have been used previously to cluster many ncRNA predictions.

[Michael]

Thank you Paul. I ran the search via the web-site. I know they use a blast pre-filter for speed. I have also tried to run infernal on the CM directly but this takes ages to complete. CMfinder looks very promising too, hope you don't mind i put in some links.

[Paul Gardner]

No prob. I would've added links but last time I tried here I was blocked.

Answer 3

[Jaydon]

BTW, we also do 3' and 5'RACE to confirm the validity of the ends. And you'll be surprise to find alternative splicing variants for most of long candidates.