I've recently run two searches, one with blastall (from legacy blast) and one with modern blastn. Both versions are 2.2.25 (the most recent).
The blastall search is: blastall -i test.fasta -p blastn -d old_formatted_database -m 8 -e 1e-10
The blastn search is: blastn -query test.fasta -db new_formated_database -outfmt 6 -evalue 1e-10
Note that the two databases have each been created with the program's respective formatdb utility.
The blastn search yields 3 results, while the blastall search results in over 50. Why is this? I realize the programs themselves are not readily comparable, but I was under the impression that their output should be similar since they're based on the same scoring methods.
From a more theoretical point of view, shouldn't there be an exact number of hits matching a given query sequence and scoring scheme? Shouldn't these all be reported regardless of the blast approach used?
can you upload and share your data for reproducing this (meaning both databases and the test.fasta), or if not appropriate make another test.db? Otherwise, I would guess or hope that there are some different parameter defaults to explain this. Also, what happens if you use a query identical to the blastall query but using the legacy command wrapper in blast+?
I can't upload the data, but I'll make a test DB later today. I did try running the query in the legacy wrapper, and I got a 3rd, different set of results (over 150 hits). I'm also curious about the defaults; I suspect this is where the difference arises. I was hoping someone here might have some insight into the differences in these defaults.
Was the query particularly short or long? Or the database huge? Because BLAST+ will automatically alter "-task" to enable megablast or blastn-short modes instead of regular blastn. Whereas BLAST has megablast as a separate application.
Also, low complexity filtering could be different between BLAST and BLAST+, limiting search results in one of them. Use "blastall -F F" and "blastn -dust no" to compare properly.
Any news on the test-case?