There are many computational miRNA target prediction tools like Diana-MicroT, TargetScan, Targetminer etc. They all provide some kind of prediction score(s) depending on target site features (e.g. seed complementarity, sequence conservation, free energy of binding, AU content, accessibility of binding sites etc.) for each predicted miRNA-mRNA interaction. However, I have read some papers that when they integrate the prediction results of these tools for their miRNA of interest, they tend to ignore prediction scores and just look at if a particular miRNA-mRNA interaction is predicted by at least 2 or 3 prediction tools they selected to assume that the miRNA-mRNA interaction is reliable. Why? I don't understand.
My questions are
1. Is it Ok to ignore prediction scores and consider overlapping results only to conclude a miRNA-mRNA interaction is reliable if we are looking at/integrating the prediction results of, let's say, 2 or more prediction tools? Or
2. Is there any tool that integrates prediction results of some of the prediction tools and their scores in some way and provide single score to rank miRNA-mRNA interactions?
Thank you all for your answers.
Hi
Very interesting topic... In fact the use of several prediction tools together has been already integrated by applications such as Mirecords (the server is not working), Mirwalk and MirDIP. You can see my webpage for details:
https://sites.google.com/site/mirnatools2/multiple-predictors
Choosing targets predicted by 2-3 applications is just a possible option. I use to choose the targets which are predicted for more than 50% of the applications that you use, meaning a bigger strength in the prediction, nothing more.
Unfortunately, scores of each application are calculated in a different fashion, so you cannot compare them directly. I think that the most reliable approach is the one described by MirDIP application, however I am afraid that the databases used for this application are slightly outdated.
I am open for further discussions. Hope it helps.
Best
Paco
@Francisco: Actually, I was aware of your webpage even before posting the question when I read your answers to other miRNA related questions posted by other members. Very informative page, thank you. I learned a lot from there. I am now checking how the tools you mentioned work.
@Ka-Chun Wong: Thank you for your answer. Picking top K predictions of each method after ranking seems good idea I understand. But there are always people who needs justification for choosing any number of K, why not N for example. Do you have any idea about that? Also, the second link you provided is giving "... can't find the page" notice. Was it working with you? Can you repost it? It looks very interesting.
Yes, there are certain limitations of these miRNA prediction tools. In our work, we tried to combine both sequence- and structure-based properties to enter into conclusion. Moreover, we also showed that accounting such parameters, you may be able to generate a mathematical model which can tell you about your predictions.
We also gave evidence about the combination of scoring functions may enhance the
prediction of (non-conserved) mature miRNAs from the (rice) genome. Go through this text:
http://www.tandfonline.com/doi/abs/10.1080/1062936X.2013.875941?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%3dpubmed&#.U84mWOOSzKU
@Prasanth: Can you upload/attach the copy of the file of your paper here if possible?
The Bioinformatics paper that Andreas mention is, from my point of view, quite important. A main conclusion is that you should NOT use combined predictions at all. My advice is to run you analyses with different prediction algorithms separately.
Dear Bahtiyor,
Considering the huge amount of false positive results that come out of prediction algorithms, one cannot just simply say that because a few of these algorithms give the same results, the result is in fact a true prediction. As for the prediction scores you would have to inspect how each and everyone is calculated and see whether or not you could somehow normalise them before you attempt to use such scores together. There are also other methods to narrow down the huge number of target predictions that come out of the many algorithms available. Some of which have been mentioned here:
http://www.researchgate.net/publication/258856319_Computational_Methods_for_MicroRNA_Target_Prediction
Good Luck
Article Computational Methods for MicroRNA Target Prediction
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biomolecules
- Nucleic Acids
- RNA
- Small RNA
- microRNA