Hi all, my first post. My day job is quantum mechanics and computational chemistry on proteins, however, before I'm able to apply either I'm facing a bit of a bioinformatics issue and I require some more seasoned advice.

I am trying to build a homology model from a protein sequence that has yet been catalogued. Through Schrodinger Maestro a BLAST and PSI-BLAST search using its default settings return poor bit-scores. Moving aside from Maestro, I decide to try PSI-BLAST via the NCBI site and via the EMBL site.

Both sites yield different hits, with the NCBI PSI-BLAST throwing up PSSM warnings after the second iteration and showing a very poor bit-score on the top hit after the second iteration. Yet, the EMBL site will happily iterate to a third step yielding some good results e.g., bit-score > 180 (if not poor overall coverage, but where it does cover the identity is approx 40% and similarity approx 50%). What could be the result of the seemingly same tool returning two different sets of results?

Looking through the submission details on EMBL, the settings are: --debug --matrix BLOSUM62 --opengap 11 --extendgap 1 --exp 10.0 --expmulti 0.001 --scores 500 --numal 500 --align 0 --dropoff 15 --finaldropoff 25 -F F --database pdb --preselids

Looking through the submission summary on NCBI, the settings are:

Program blastp <--- threw me at first, but it is PSI-BLAST

Word size 3, Expect value 10, Hitlist size 500, Gapcosts 11,1, Matrix BLOSUM62, Filter string F, Genetic Code 1, Window Size 40, Threshold 11, Composition-based stats 2, Database pdb <---I added this entry in myself, it is not mentioned up in the summary but it was selected>

Per input entry screen, I made as many of the settings as identical as possible, although EMBL's implementation of PSI-BLAST has more parameters to change e.g., dropoff, final dropoff, alignment view, usage mode for PHI-BLAST. Further, the EMBL implementation states:

Position Specific Iterative BLAST (PSI-BLAST) 1) refers to a feature of NCBI BLAST 2.0

And the NCBI PSI-BLAST states:

Program: BLASTP 2.6.0+

I would appreciate any thought as to why I am seeing a different set of results?

Many thanks, Anthony