Biopython Ncbistandalone Blastall Gives Different Result Than Calling Blastall Directly From Cmd

Question

[Niek De Klein]

So, first I tested what results I should get from the blastall program using the command line, with e-value 0.001:

C:\Niek\Test\blast-2.2.17\bin\blastall -p blastp -d C:\Niek\Test\arabidopsis-smallproteins.fasta -i C:\Niek\Test\arabidopsis-HD.fasta -e 0.001 -F F -m 8 -o C:\Niek\Test\arab-HD-smallproteins-notfiltered.out

and

C:\Niek\Test\blast-2.2.17\bin\blastall -p blastp -d C:\Niek\Test\arabidopsis-smallproteins.fasta -i C:\Niek\Test\arabidopsis-HD.fasta -e 0.001 -m 8 -o C:\Niek\Test\arab-HD-smallproteins-filtered.out

After that I made a local blast program, which works fine but it only found 91 results with e-value equal or lower than 0.001, where the results from the blastall via cmd gave around 140~ something results. I first thought it missed some, but all the e-values are different.

from Bio.Blast import NCBIStandalone
from Bio.Blast import NCBIXML

my_blast_db = r"C:\Niek\Test\arabidopsis-smallproteins.fasta"
my_blast_file = r"C:\Niek\Test\arabidopsis-HD.fasta"
my_blast_exe =r"C:\Niek\blast-2.2.17\bin\blastall.exe"

result_handle, error_handle = NCBIStandalone.blastall(my_blast_exe, "blastp",
my_blast_db, my_blast_file)

blast_records = NCBIXML.parse(result_handle)

E_VALUE_THRESH = 0.001
x = 0
for blast_record in blast_records:
    blast_record = blast_records.next()
    for alignment in blast_record.alignments:
        for hsp in alignment.hsps:
            if hsp.expect <= E_VALUE_THRESH:
                print "==========Alignment========"
                print "sequence:", alignment.title
                print "length:", alignment.length
                print "e value:", hsp.expect
                x += 1

I first thought that the local blast from biopython uses a different algorithm, but at 'myblastexe =r"C:\Niek\blast-2.2.17\bin\blastall.exe"' I specify the same program but it should be the same. Then I thought it had something to do with the filtering option but I checked both filtered and unfiltered and wasn't any of that.

If you know why the local blast from biopython NCBIStandalone gives a different result than doing it directly in the cmd, please let me know.

Thanks in advance, Niek

edit: I checked, and it seems that the NCBIStandalone filters out 100% identities, which the blastall called by cmd does not. However this doesn't explain why the e-values are so different.

[brentp]

i'm not a regular windows user, but is it a typo that your command-line blast paths don't end in .exe while the one you send to biopython does?

Answer 1

[Peter]

One major problem is you are skipping half the results with this:

for blast_record in blast_records:
    blast_record = blast_records.next()
    ...

It should be just:

for blast_record in blast_records:
    ....

Or if you would rather call the next method explicitly for some reason, something like:

while True:
    blast_record = blast_records.next()
    if blast_record is None: break
    ...

Secondary issue: blastall is now being phased out by the NCBI who call it "legacy" BLAST, they encourage people to use BLAST+ instead, in this case blastp at the command line. As a result, the old Biopython wrappers for calling "legacy" BLAST are all considered obsolete.

[Niek De Klein]

Thank you for the answer!