Tallying BLAST Matches

[JuliaEpps]

Hi,

I've got a custom BLASTDB created from a FASTA file containing large amounts of short sequences:

>seqA
GATCGATAGTTAGACTAGTAAGCAAA
>seqB
AAATCACAGTGACTAGTAGAGAGATT
>seqC
AAGGCCCCTATATAGACTGACTAGTA

and so on.

I've sequenced a large library of molecules which are ligation products between 2 of the sequences held in this custom DB i.e. SeqA-SeqC (GATCGATAGTTAGACTAGTAAGCAAAAAGGCCCCTATATAGACTGACTAGTA).

I need to be able to BLAST my sequencing data against the custom database and then tally how many times each possible ligation product is present and output this based on their FASTA labels i.e. something like this:

SeqA-SeqC 54
SeqA-SeqB 102

Is this something that BLAST/BLAST+ may be able to do instrinsically or is it something that would have to be done in perl/python? Does anybody have a script which may do something similar which I could have a look at to get a framework to build upon?

Thanks!

- Julia

[JuliaEpps]

I thought of using -max_target_seqs 2 on command-line blastn providing the two top-hits which will represent which 2 sequences gave rise to the ligation products and then outputting this to a custom BLAST format - I am not sure how to proceed from there to obtain the required tally.

[gringer]

This mostly depends on what you mean by "large amounts" and "large library".

If there are fewer than ~100 short sequences, you could probably get away with enumerating (and indexing) all possible combinations (10,000 total), then doing a simple search.

If you have fewer than ~100,000 short sequences and have predictable start/end points in the library, then you could hash the short sequences and store in memory, then check each ligation product part against the hashed index (unfortunately, using hashes wouldn't allow for errors).

Otherwise, if the library is less than around 1-5GB, you might be better off reversing the query -- create an index of unique ligation products and map the short sequences to those.