How to solve the different number of hits from tblastn when I use the option -parse_seqids in makeblastdb command?

Question

[carina2817]

Hello,

I am running tblastn and I am getting different hit numbers depending on if I add the option -parse_seqids to makeblastdb .I initially ran the following command to make the database:

makeblastdb -in GCA_003024985.1_Erow_1.0_genomic_Euperipatoides_rowelli.fna -dbtype nucl

then I ran tblastn with the next command:

tblastn -db GCA_003024985.1_Erow_1.0_genomic_Euperipatoides_rowelli.fna -query S_cerevisiae_all_prot_uniq_join.fa -out S_cerevisiae_all_prot_E_rowelli_tblastn_ful_test_original.out -outfmt '6 qseqid qgi qlen sseqid bitscore length pident qcovs evalue qstart qend sstart send qseq sseq' -num_threads 20

The output has 194 hits.

But then I was trying to retrieve the hits DNA sequences from tblastn to run the reciprocal blast test and I had an error which I solved adding the option -parse_seqids to makeblastdb [https://github.com/lindenb/jvarkit/issues/134]:

makeblastdb -in GCA_003024985.1_Erow_1.0_genomic_Euperipatoides_rowelli.fna -dbtype nucl -out blastdb_E_rowelli -parse_seqids

and I ran tblastn again with this new database:

tblastn -db blastdb_E_rowelli -query S_cerevisiae_all_prot_uniq_join.fa -out S_cerevisiae_all_prot_E_rowelli_tblastn_ful_test.out -outfmt '6 qseqid qgi qlen sseqid bitscore length pident qcovs evalue qstart qend sstart send qseq sseq' -num_threads 20

This time I am getting 195 hits. I am working with many genomes and I have this problem with some of them (the biggest difference is 20 hits). Do you have any idea how to correct this or which output I should select?

Thanks.

Paola

[gb]

Can you confirm that makeblastdb -parse_seqids does not give a warning like "duplicate accessions found" or something like that

[carina2817]

I don't get any warning message

Answer 1

[carina2817]

It comes out the problem was not the "-parse_seqids" option, I discovered that every time I ran blast with some of the genomes I am using the results file had a different number of hits and this happens (with some genomes) in all blast versions after 2.3.0. I sent an e-mail to blast support and they told me that's a new bug and that they would make a report, the problem is produced when using many processors, using 1 processor does not produce the problem and the results must be consistent at some number of processors above 1.

[gb]

oh! thanks for updating your post