CircRNA discovery tools

[warrenemmett]

Hi,

Circular RNA has become a hot topic recently, I know there are several papers but none that I've found that include the software pipeline they used.

If anyone is willing to provide a tool they are creating I'd be happy to test it or know of a any available scripts I'd be very grateful!

[Wallysb01]

I don't think any pretty tools exist yet. I've been looking into this recently myself and it looks like people have coming up with rather laborious in house pipelines.

Check out the recent nature paper (Circular RNAs are a large class of animal RNAs with regulatory potency, http://www.nature.com/nature/journal...ture11928.html) for example. In the the "Computational pipeline for predicting circRNAs from ribominus sequencing data" section of the methods they have a few paragraphs and a couple of example commands describing what they did, but to recreate that in a more easily reproducible pipeline would take some time, at least for me.

[warrenemmett]

Hi there,

Thanks for the reply. Yes I saw their paper and came to the same conclusion. The nail in the coffin is that they do not include any of their scripts which essentially means redoing most of their work.

[Wallysb01]

Hi there,

Thanks for the reply. Yes I saw their paper and came to the same conclusion. The nail in the coffin is that they do not include any of their scripts which essentially means redoing most of their work.

Exactly, I wonder if they are trying to get a two-for-one and we'll see a methods paper from the same group in the near future? For bioinformatics it would be nice is editors and reviewers required publishing scripts require for doing the analysis in the paper, at least if it is something substancial. I'd sure like to look for circRNA in my data sets, but I'm not sure I have the hours to recreate the required tools.

[Palgrave]

I got an email from the authors. They are writing a methods paper that is going to be published in near future (weeks). Stay tuned for that. I guess we could use the BED-file they provided in the meantime. At least if you are working with human, mouse or c.elegans.

[GenoMax]

I got an email from the authors. They are writing a methods paper that is going to be published in near future (weeks). Stay tuned for that. I guess we could use the BED-file they provided in the meantime. At least if you are working with human, mouse or c.elegans.

Even if authors are writing a paper they could make the scripts/programs available (at least to people like you who ask along with an appropriate disclaimer).

[nikolaus_rajewsky]

Dear All,

Thanks for your interest in circRNAs. I am the corresponding authors on the Nature paper about circRNAs you are mentioning, and I wrote the email to "Palgrave". I would like to clarify the contents of my email to him/her:

-- I wrote to him/her that we right now hesitate to send out the scripts because they are not commented/cleaned up properly, and we got a lot (!!) of requests.

-- We thus decided to send them out to anyone who asked once they are cleaned up. This will be the case in a few weeks when we will submit a new paper. I am saying (and wrote in my email) submit, not publish.

-- In this new paper we apply these scripts to validate our novel unpublished biochemical protocol for the purification of circRNA that we developed and to compare yet to another method. thus i would hardly call this a "2-for-1".

thanks again for the interest

Nikolaus Rajewsky

[Wallysb01]

-- In this new paper we apply these scripts to validate our novel unpublished biochemical protocol for the purification of circRNA that we developed and to compare yet to another method. thus i would hardly call this a "2-for-1".

Its good you're doing more, and it sounds like much more, than just publishing your scripts from your previous paper in the upcoming protocol paper. However, if you're creating a whole new method for analyzing data that lead to the major findings in your paper, you should really have released the those methods with the original paper.

This isn't just your fault, this is the reviewer's and editor's fault as well. Code used in research needs to be made available with the release of the papers using it. And the whole field of biology needs to figure this out. It is critical in reproducing findings and advancing the field in general. You sound like you're at least going to make this available eventually, but by not being required to release it with the findings created by it, it would be quite easy for it to never be publicly available.

[lewewoo]

I think some people (or all) should mention this issue to Nature editors as comments or letter to editors. It is important to include codes in the paper upon published. All high end journal requests raw sequencing data and processing steps, therefore, it is reasonable to release the codes with original paper together. Editors should set up rules for this. I think this discovery is very cool and many people want to check their own data (including me). The more people write to the editors, the more chances for them to responses. For examples, many people in a Chinese biological forum have concerns on a Cell paper about faking data, they sent emails to the editor of Cell to ask author providing original data, eventually, Cell editors responded and the author retracted the paper (they did fake some data). I am not saying the data of circular RNA paper has problems, (the discovery and the paper is great and awesome!) I am saying that if we all send emails to require Nature editors to do something for setting up scripts-sharing rule for publishing paper, Editors will notice this issue and in the future, more people will benefit from the great discovery.

[rnaNGS]

decoding miRNA-circRNA interactions from CLIP-Seq data

starBase v2.0 have provided miRNA-circRNA interactions by scanning circRNA sequences overlapping with CLIP-Seq(PAR-CLIP, HITS-CLIP, CLASH, iCLIP) peaks for potential microRNA targets (mirSVR/miranda) and then output the detail informations. You can browse and download these miRNA-circRNA interactions in our starBase website (http://starbase.sysu.edu.cn/mirCircRNA.php).

We can identify 52 CLIP-supported miR-7 target sites in the circRNA CDR1as reported by Memczak et al. (Nature, 2013, 495, 333-338. ) and Hansen et al. ( Nature, 2013, 495,384-388.).

[ashish_yeri]

CircBase has the scripts by Rajewski required to detect circular RNA and also some custom python scripts for filtering the results. Has anyone understood the field named anchor_overlaps in the sites.bed output? Any help will be greatly appreciated.

Thanks,

[NKAkers]

I'm also trying to parse through the circbase output. I only seem to get values of 0, 1 or 2 for anchor_overlap, so it seems to be a classifier, not a count value as I initially suspected.

If anyone has info on the rest of the custom headers in sites.bed/circ_candidates.bed I would love to hear it.

[NKAkers]

ok ashish_yeri, I think I've figured it out.
anchor_overlap refers to this point from the nature paper:

"4) the breakpoint cannot reside more than 2 nucleotides inside an anchor; "

so anchor_overlap is the number of nt in from the edge of the anchor where the circRNA breakpoint is.
The only part I'm still confused about is the best_qualA/B fields. The filter they implement is:
"(6) unique anchor alignments with a safety margin to the next-best alignment of at least one anchor above 35 points (~more than two extra mismatches in high-quality bases); "

But the circbase code requires only one of best_qualA/B be >35. Without knowing what exactly these scores are, it's hard to know what I'm doing.

[ashish_yeri]

Dear NKAkers,

Thanks for your reply. If the splicing signal is found within the anchor, the anchor_overlap refers to the number of nucleotides from the anchor flank. The maximum is 2 like you point out. The best_qualA refers to the difference between the best and second best mapping quality for anchor A. If this value is 0 or a very small one, then that would mean there are at least 2 high quality mappings for anchor A, which would make it ambiguous. Thats why we're interested in only those reads for which the difference is > 35. The same holds good for anchor B. For a more stringent filtering scheme, I would filter both for anchor A and B. Hope that helps!

[mswierniak]

Hi,
I wonder if you have found the reasonable way to filtering circRNAs. Filtering proposed by authors seems to be very liberal so if I'm interested in searching for "the best" of potential circRNAs the filtering thresholds should be more restrict. The way I'm thinking about filtering is to use the number of unique reads and the number of reads but I'm still not sure what exactly means "unique reads". If they are unique regarding their sequences or these anchors mapped uniquely. Did anyone try to do some kind of normalization of these reads to estimate expression of potential circRNAs?