miRDeep2 issues with non-model organisms

[dawndee63]

I'm having trouble with miRDeep2 results on a non-model organism that has no known microRNAs. Tried using mature microRNAs from a closely related species, but all predicted microRNAs show a Randfold p-value of 'no,' suggesting theyre not real. My pipeline worked fine on another organism with known microRNAs , so i'm not sure what's wrong this time. Any advice?

 

[Akeem]

It seems the reference microRNAs might be too far off, or your genome assembly's not tight enough, I'll suggest you try tweaking Randfold thresholds, or even train the model with de novo candidates first, it could help flag legit hits

 

[TropicalMe]

When working with a non model species with no known microRNAs, using references from related species can help but often isn't perfect. The Randfold 'no' p values usually mean the predicted hairpins aren't statistically significant, which could mean your input data or parameters need tweaking

 

[st3@k to_go]

I'm having trouble with miRDeep2 results on a non-model organism that has no known microRNAs. Tried using mature microRNAs from a closely related species, but all predicted microRNAs show a Randfold p-value of 'no,' suggesting theyre not real. My pipeline worked fine on another organism with known microRNAs , so i'm not sure what's wrong this time. Any advice?

Another thing to consider is the size of the genome you're mapping to. If it's a large, complex genome, it might just be finding more random hairpin-like structures. You could try relaxing the Randfold stringency slightly, though that's a bit of a last resort

 

[Josef_10]

I'm having trouble with miRDeep2 results on a non-model organism that has no known microRNAs. Tried using mature microRNAs from a closely related species, but all predicted microRNAs show a Randfold p-value of 'no,' suggesting theyre not real. My pipeline worked fine on another organism with known microRNAs , so i'm not sure what's wrong this time. Any advice?

Yeah, @TropicalMe is correct. That can happen when working with non-model organisms. Anyway, the 'no' p-value means the predicted hairpin structure isn't stable enough, which is a key characteristic of a real pre-miRNA. Essentially, the sequences are failing the secondary structure check, which is a crucial step in microRNA annotation.

Try relaxing your filtering stringency as suggested and see if you get any hits. Maybe try a broader search for microRNAs from a whole family of related organisms, rather than just one species. You could also try to de novo predict microRNAs without using any known microRNAs as a reference. This is a bit more of a shot in the dark, but it can sometimes yield results. Just be prepared to validate any hits you get through additional experiments.

I think this is a good case showing that microRNA prediction in non-model organisms is a challenge. Sometimes, the 'no' p-value really is telling you there are no good candidates, or at least none that fit the strict criteria of the program. It's a game of troubleshooting and iteration, but keep at it!