Psiblast Warning: Composition-Based Score Adjustment

Question

[Stef]

Warning: lcl|Query_1 1DCH:A|PDBID|CHAIN|SEQUENCE: Warning: Composition-based score adjustment conditioned on sequence properties and unconditional composition-based score adjustment is not supported with PSSMs, resetting to default value of standard composition-based statistics

Can someone help me as to what this warning message means? I mean the results that I get are pretty fine, so until now I just ignored it. But slowly I'm wondering if I'm doing a mistake.

I run PSI-Blast with following parameters

./psiblast \
  -query 1cbyA.seq \
  -db ./blastdb \
  -matrix PAM30 \
  -num_iterations 5 \
  -num_alignments 3000 \
  -num_descriptions 3000 \
  -gapopen 32767 \
  -gapextend 32767 \
  -evalue 10 \
  -outfmt 7 \
  -out ./output.txt

[crystalrose]

Hello,Stef!I'm facing the same problem as you met 3 years ago. Did you find an appropriate method to solve it?I just set the parameter -numiteration to be 2,but still the warning appears.

[ChuanD]

I meet the same problem. Did you solve it?

[ChuanD]

I added "-comp_based_stats 1" option, which avoid to print the warning information.

Answer 1

[Michael Kuhn]

Composition-based score adjustment adapts the substitution matrix to the context of the sequences to increase performance, see Yu & Altschul, 2005. PSI-BLAST, on the other hand, generates profiles for each position, so there is no substitution matrix to adjust after the first iteration. Therefore, the warning makes sense, but you don't need to worry about it.

I'm not sure why the warning is triggered, your params seem fine (although I can't judge why you force an ungapped alignment, this seems to contradict the aim of finding more remote homologs with PSI-BLAST).

[Stef]

Hm but I do give him as parameter a substitution matrix. If it's only used for the first iteration I guess in the second it throws the warning or? So that would be the reason(still kind of stupid to throw it though)

Well for my application I can only use ungapped alignments so I have to forfeit the advantages of finding remote homologs with gaps since the are useless to me.

I mean if another application exists that can do fast sequence alignments without gaps I guess it would fit my needs better, but I don't know of any, so I just force BLAST and HHSearch to no gaps.

Thanks