miRDeep2.pl cannot recognize reference fasta file

Question

[realnewbie]

Hi,

I have been trying to run mirdeep2 for a while. However, the tool cannot recognize bowtie-indexed reference genome at mirdeep2.pl step. It gives the same error: "refence.fa is not a fasta file". However it runs same file for mapping. When it comes to mirdeep2.pl, it could not. I could not understand why. Even little help will be appreacited a lot. Thanks a lot!

[GenoMax]

Are you using plain fasta file? You can't use bowtie indexed genome.

[realnewbie]

What is "plain fasta" file? I have been using this: Homo_sapiens.GRCh37.dna.primary_assembly.fa. In tutorial, it says we need to index the fasta file first.

For novel miRNA prediction we need to map the reads against a reference database which has to be indexed by bowtie 1. For this we take a reference database file, lets call it refdb.fa (This can be a genome file or simply a file with scaffolds) and build a bowtie index by typing

From this site: https://drmirdeep.github.io/mirdeep2_tutorial.html

[GenoMax]

If you followed instructions in the tutorial I suppose you used

bowtie-build Homo_sapiens.GRCh37.dna.primary_assembly.fa Homo_sapiens.GRCh37.dna.primary_assembly.fa

Here Homo_sapiens.GRCh37.dna.primary_assembly.fa should be the original fasta format genome file. Are all result files from bowtie-build command in the same directory?

[realnewbie]

Yes, they are in the same directory. By the way, thanks a lot for your quick response to help.

[realnewbie]

Tutorial page sucks. It was not useful at all. It was impossible to download the relevant files properly without conda install.

[GenoMax]

You may want to follow this tutorial. The files referred to there are probably in the tutorial directory in miRdeep2 installation.

[realnewbie]

Thanks a lot! I realized that in the mapper we are using bowtie indexed reference genome, however, in the mirdeep.pl step, we need to use not indexed-original fasta file. The old tutorial was a little bit confusing regarding this issue. New tutorial is more helpful. Best,