Aligning sequences against RFAM and filtering them out that are not aligned to miRNA.

for details check these links this and this

Assuming you have a reference genome for your organism, one option is not to worry about pre-filtering the miRNA seq data. After adapter trimming and alignment to the reference genome, rRNA and tRNAs that produced spurious small RNAs can be recognized easily in at least two ways --

1, based on the existing annotations of rRNA and tRNA genes in the organism, such loci can easily be found.

2, more universally, after reference-mapping, small RNA clusters that are NOT dicer-derived will often have a distinct distribution of read sizes that is NOT within the normal ~20-24 nt range. In contrast, the miRNAs you are interested in will produce clusters tightly centered around 20-22 nts.

For aligners, probably something faster then blastn is appropriate .. bowtie1 is a great aligner for small RNA-seq data.

The tRNA thing is actually a valid fear - you will get mismappings to miRNAs unless you allow 0 mismatches. In fact there are a couple nucleotides that are added onto tRNAs that can make them not fully map to their genomic precursors but actually look even more like miRNAs.

I would download all tRNAs and use them as a blacklist: http://gtrnadb.ucsc.edu/download.html (rfam is probably even a better resource)

So before aligning to the genome you should align to the tRNAs, and remove those sequences that match perfectly.

miRNAbase is a good database for miRNA research. You may want to do homology search first. Also. you can do target prediction of your miRNAs.