Criteria for novel miRNA

[hamid-ali-khan]

Hi all,
1:can some body tell me that what the minimum read count (mature) for novel miRNA ?
2: what is the estimated probability that the miRNA candidate is a true positive?
3: if the p value of a miRNA is non significant, can we still consider it ?

[mbblack]

Hi all,
1:can some body tell me that what the minimum read count (mature) for novel miRNA ?
2: what is the estimated probability that the miRNA candidate is a true positive?
3: if the p value of a miRNA is non significant, can we still consider it ?

1. No idea - that is like asking what is the minimum read count that any sequenced transcript is real. Under ideal circumstances, even a single uniquely and perfectly mapped read may be considered real. Most people would likely feel more comfortable with a higher cutoff, and for RNAseq differential gene expression, a number of published papers have used arbitrary minimum count cutoffs of 10 (but I have also seen 5 used in at least one paper).

2. That will be determined by your particular experiment. Estimation of true positives or negatives can only be done relative to the experiment in question. You will have to compute your own statistical estimates for that.

3. Any p-value cutoff is purely arbitrary, even the most often used "p-value < 0.05". So you can adjust your selected cutoff up or down as you deem necessary and in light of your own experiment. All the p-value does is tell you something about the likelyhood of that event occurring relative to chance - it says nothing intrinsic about the biological reality nor relevance of that event. That interpretation is up to you.

Often people follow up on statistically non-significant data based on their or published previous work on it and their interpretaiton of the biological importance or relevance of that data. Or, sometimes (human clinical work for example) you do not have enough samples for robust statistical analyses so have to relax your cutoff criteria for significance. As long as you have logical reasons for your decisions and can defend them, you can proceed as you decide.

That is the big caveat to 3. Should you report and/or emphasize a statistically non-significant miRNA, you need to have a rational validation of why you are presenting it. If the statistics tell you that it is highly likely to have occurred purely by chance, then you need to be ready to defend why you feel the statistics are not relevant to the inclusion of that data point in your results.