I wonder if I use 3 online programs for mir prediction, will the results showe the corresponding similarity only 1/3 or 2/3 among 3 programs? Is it significant? How?
Usually in bioinformatics, each tool has an underlying statistical model to find a pattern, and the tool will present results which pass a threshold for statistical significance.
If two tools use a different type of statistical model then they will find two different sets of results, with some overlap. If two tools use a similar model, but with different thresholds of sensitivity, then one method will find results that are a subset of the other method.
If you use three different tools, and plot a Venn diagram of the overlapping results of the three programs, you will find out which predictions are 'robust' such that they are found by multiple methods, and you will find which methods are more sensistive and which have higher specificity.
Hope this helps.
There are also different types of microRNA prediction tools. Some are based on conservation and some are based on small RNA-Seq data. If you compare these two, there will also be a lot of differences, since these based on conservation cannot find species specific microRNAs and these based on NGS will not find microRNAs which are not expressed in the sample. But if you compare tools of the same group, there should be quite a big overlap, since all of them use the same ideas (conserved region/expression and secondary structure).
To answer your question: Do what Jonathan recommends! Test and visualize it.
If you have NGS data, I would recommend miRanalyzer and mirDeep.
If it is NGS data, you can also use iMir that uses both miRanalyzer and miRDeep2 to predict novel miRNAs.
Check out http://www.nature.com/natureevents/science/events/47533-microRNA_Analysis_Using_Next_Generation_Sequencing
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biomolecules
- Nucleic Acids
- RNA
- Small RNA
- microRNA