I don't know any good tool for ncRNA prediction without a reference genome.

BUT:

You can try to align your reads against a reference genome of a species that is evolutionary close. ncRNAs are well conserved and the chances are high that the reads also map to that genome. And then you can use several ncRNA prediction tools.

For microRNAs, you could start with the most frequent sequences in your short RNA-Seq dataset. When counting and ranking the reads, normally microRNA sequences rank at the top (together with low complexity sequences, or PCR clones). You could pick out the top 1k and align these sequences against a microRNA-set (from miRBase) of an evolutionary close species (microRNAs are very well conserved). This way, you can find microRNAs in you new species without a reference genome.

Thank you David. But I would like to disagree with you small ncRNAs are not much conserved. Almost all species possess some small ncRNAs specific for them. This is the main reason we are still not able to identify the complete repertoire of small ncRNAs. I agree on your strategy of aligning the small RNA seq data to known ncRNAs but there are two problems with that approach, Firstly, the reads are very small in length and they can align ambiguiously to any random position. Second, even if the reads aligned unambiguously, then too I will have information only about the conserved small ncRNAs while I wish to identify some novel or species -specific ncRNAs too

I don't wanna sound smart-alec, but the majority of small ncRNAs are highly conserved. Nevertheless, I have to agree with you that species-specific small ncRNAs exist and are always hard to locate.

With your second assumption that the reads are very small in length and they can align ambiguiously to any random position, I have to disagree. If that would be the case, the complete technique of short RNA-Seq would be useless. The reads of functional ncRNAs are highly not-random and most of them can be mapped to their ncRNA of origin. You might map them to e.g. microRNAs of the same family, but that shouldn't be a problem. Here I'm talking about reads directly mapped to ncRNA sequences (not a genome).

Your last thesis, that you will find only known ncRNAs is correct... but only if you map them to a set known ncRNAs of a close species. My suggestion was to map the reads to a genome of a close species. That is a difference, because this way you might indeed find new ncRNAs!

But in the end, the only way to predict new short ncRNAs with short RNA-Seq data is to make a second experiment. Sequence not only short RNAs (18-30nt), but also a longer fraction (30-100nt). This way, you also sequence not only the mature sequences, but also the precursor of the ncRNAs. Then you can map your short reads against the long ones and predict short ncRNAs.

I hope that information helps you a bit. :)

Yes it really helps a lot thank you David

I am not well versed with small RNA sequencing. But I can suggest the following

- Use miRDeep2 (https://www.mdc-berlin.de/8551903/en/research/research_teams/systems_biology_of_gene_regulatory_elements/projects/miRDeep) and mirTools (http://centre.bioinformatics.zj.cn/mirtools/) to detect known and novel small miRNA from the dataset.

- Further you can run miRFinder (http://www.bioinformatics.org/mirfinder/) on the genome to predict all miRNA precursors and compare the known and novel sequences with the location of the predicted precursors.

- A final approach can be to consider only those novel candidates which are predicted by the three softwares, hence reducing putative false positives.

- As for the identification of possible conserved miRNA sequences my guess is the miRanalyzer (http://bioinfo5.ugr.es/miRanalyzer/miRanalyzer.php) tool should work fine.

If you prefer GUI rather than breaking head at the linux console try the smallRNA workbench (http://srna-workbench.cmp.uea.ac.uk/tools/). Howver apart from the tools I would encourage you manually curate your predictions on a genome browser and associate expression of known miRNAs with what is mentioned in the available literature. happy small RNA hunting!!

Hi Migraya,

I hope you solved the problem. However, if you dont, I would like to invite you to send us two FASTA files, one with reference ncRNA sequences (not more than 100) and the other with false ncRNAs that you found be confusing in your work (between 50 and 100), to see if the Universal Feature Method (UFM) we have developed is able to give a useful sign for classifying such kind of small sequences.

Regards

Diego Frias (Brazil) 

Dear Diego,

Thank you for your response but I have solved problem of annotation.