Is there a method or HMM set to quickly determine if a sequence has a tRNA sequence for a large number of sequences/reads?

Question

[O.rka]

I am looking at over 10 million reads in the form of fasta. I want to determine if each sequence has a portion of a tRNA sequence. I've been using tRNAscan-SE but this takes a very long time for a large number of reads since it was designed for contigs.

Scanning for HMMs is very quick. Is there a tool that does this or a HMM set that I can use to say "read_X has or does not have a tRNA hit"

I looked in the source code of tRNAscan-SE and didn't notice any HMMs.

Answer 1

[Mensur Dlakic]

tRNAscan-SE uses covariance models, which one can think of as HMMs that simultaneously model and score both primary sequence and secondary structure. This is why tRNAscan-SE is slow, but that's the price of extra calculation to get maximum sensitivity.

The fact that tRNA requires base-pairing between relatively distant portions of the sequence means that any tool scanning short sequences will miss secondary structure interactions. Still, there seems to be one:

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0077596

It includes a covariance model for tRNA, but be warned that it will not be speedy. I suspect that any increase in speed will be at the expense of specificity.