Not getting all hits in blast+ all. vs. all

Question

[bioPiraten]

I realize there are multiple threads and extensive documentation about blast+, but after traversing most of that information, I haven't been able to find the answer to my problem.

I am making an all vs. all blast+ search on private makedb database.

I got 529 proteins in my DB, which I make by:

makeblastdb -in C_domains.faa -dbtype 'prot' -out Cdomains_DB

Then I blast against it:

blastp -query C_domains.faa -db Cdomains_DB -num_threads 16 -out C.blast -max_target_seqs 529

When I parse the result I get a varying number of alignments/sbjcts hits for each query sequence varying from 529 - 400. I want to know all the alignment scores i.e. 529 alignments each time.

I tried to set -max_target_seqs 10000

but it didn't change a thing

... As a side note, I've tried the same thing for another database with 300 proteins, and it returned 300 hits each time...

Answer 1

[Michael]

Blast didn't find alignments for those sequences that were not aligned, because they are too dissimilar.

To increase the sensitivity, you can reduce the -word_size parameter to 2, and to filter out less results you can set the -evalue parameter to 1000 or so. As blast is a heuristic it might still not find an alignment for each entry. If you simply need an alignment of everything, irrespective of how bad, you should use full smith-waterman using e.g. Ssearch with very high -evalue cutoff.

Btw what do you need that for, I have the feeling, there might be a more appropriate method to achieve whatever you are trying (creating some sort of 'distance matrix'?, or multiple alignments?)

[bioPiraten]

Thanks, I will try that. I am analysing a specific protein domain, which I have extracted from 30 organisms and now I want to look at the similarity distribution, and infer phylogenetic relationships. I figured blast was the most feasible way of getting % identity for all vs. all of 529 sequences with avg. length ~300 AAs.

[Michael]

Wouldn't it be better to directly do a standard multiple alignment and phylogeny? Maybe a bit too many proteins for some msa to handle?