This is a test version of Biostars. For the public version, visit https://www.biostars.org.
stand alone psi-blast

I'm using psi-blast (blast version 2.2.30+),this is my command:

psiblast -query seq.fasta -db db -num_threads 15 -out outfile.txt -outfmt 6

I'm not sure than if I have got the right result.It seems that it only run one iteration. But if I add the parameter: -num_interations 2, warnings appear:

Frequency ratios for PSSM are all zeros, frequency ratios for BLOSUM62 will be used during traceback in composition based statistics

I don't know where is the mistake and how can PSSM are all zeros. any help will be appreciated!

software-error blast sequence-alignment
• 5,296 views • link
updated by 1992minto • 1992minto • written by crystalrose • crystalrose

enter image description herehi ! here , I get the two matrix of PSSM , but I don't their meaning .As shown in the follow picture

enter image description here

• 0 views • link
written by 1992minto • 1992minto

You have broken links. Please open a new question and add a link to this question in there so people can help you out.

• 0 views • link
written by Ram Ram

1 answer

PSSM scores might be all zeros because no domains were found in the input sequences. Could you give us more details on how the db was created?

• 0 views • link
written by Ram Ram

Thanks for your answer.I create the db using my own fasta file:makeblastdb -in seq.fasta -dbtype prot -out db -parse_seqids.The fasta file here is the same as psiblast query file.I want to classify these protein seqeuences ,which can't be classified into pfam ,by psiblast.

• 0 views • link
updated by Ram Ram • written by crystalrose • crystalrose

Are you sure running PSI-BLAST against the same set of sequences as both subject and query is a good idea? What is your ultimate aim?

• 0 views • link
written by Ram Ram

Actually, I aim to classify protein families for some protein sequences.I think I should change my strategy and search the known-function-sequence db. Do you think this can be work?

• 0 views • link
updated by Ram Ram • written by crystalrose • crystalrose

You might wanna just query against nr or SwissProt. Classification is usually done by checking homology to proteins with known function. If your query is all unknowns and you're not looking to internally cluster them, the DB should be nr or SwissProt.

• 0 views • link
written by Ram Ram

Thanks for your help!I sovled this problem by adding an parameter -comp_based_stats 0 and got some conserved results.

• 0 views • link
updated by Ram Ram • written by crystalrose • crystalrose

Log in to answer this question.

More posts like this