How to make BLASTN be aware of short read?

Question

[Xiaokang]

I'm using blastn (https://anaconda.org/bioconda/blast) to find similar sequences of a target sequence against a FASTA file. But my read is quite short (68 bases). I realised that blastn won't report any hit. But there is actually a very good one in the FASTA file after checking manually.

Here is the query sequence:

$ cat query.fa
>query read
GGAGTAGGCGCGAGCGGCAGGAGGCGGGCAGGCGGAGGGCGAGGCAGGGAGGCGCCGCCTGGAGCGCA

And this is the "good one" that I found manually (I picked it out from the original FASTA file and save it as a single FASTA file and also created the BLAST database from it using makeblastdb):

$ cat db.fa
>database read
GAGTAGGCGCGAGCTAAGCAGGAGGCGGAGGCGGAGGCGGAGGGCGAGGGGCGGGGAGCGCCGCCTGGAGCGCGGCAG

And the command that I used is:

$ blastn -db db.fa -query query.fa -outfmt "6 qseqid sseqid evalue length pident bitscore ppos"

(then no hit was reported)

But they two can actually match very well:

database           1 -GAGTAGGCGCGAGCTAAGCAGGAGGCGGAGGCGGAGGCGGAGGGCGAGG     49
                      ||||||||||||||  .|||||||||||    |.|||||||||||||
query              1 GGAGTAGGCGCGAGC--GGCAGGAGGCGG----GCAGGCGGAGGGCGA--     42

database          50 GGCGGGGA-GCGCCGCCTGGAGCGCGGCAG     78
                     |||.|||| ||||||||||||||||.
query             43 GGCAGGGAGGCGCCGCCTGGAGCGCA----     68

(above is the alignment result by needle)

Is there any parameter of blastn that I can adjust to relax the BLAST requirement cutoff?

Answer 1

[Michael]

Try the parameter -task blastn-short to adjust blast setup to short reads.

[danvoronov]

It is also worth noting, I think, that the -word_size parameter can be used. blastn uses 11 as default, but can go until 4 nucleotides, short-blastn has a default of 7 but can go as low as 4 as well. E-value cut off can also be relaxed, althout the default is 10, so it can be left blank.

[Michael]

Yes, that is very useful, reducing the word_size will also increase run time slightly. Alternatively, other aligners, like gmap, Fasta (much slower but still a bit faster than needle), or exonerate (much slower but very accurate) could be tried.

[Xiaokang]

-word_size is awesome! But blastn-short doesn't seem to have a default of 7, since -word_size 7 gave me many hits but -task blastn-short gave me nothing (with same -evalue). I only got hits with using -task blastn-short unless I set -evalue to a very high value, but then because of the high E-value, I got many noisy hits.

[Xiaokang]

Thank you Michi! I found -word_size more flexible than blastn-short though