Selection of binding site amino acids only for FMO

[haroldgrosjean]

Hello,

I am trying to setup an FMO calculation using your tool. I would like to calculate the interaction energy between a small molecule ligand and all amino acids that are 6 angstrom around it with correct neutralization/ hybridization states at the cut ends. No second layer is needed.

First I generate a setup file as indicated in the tutorial.

fragit-conf -o myconfig.in.txt
myconfig.in.txt - this is an .in file

I then use this file to generate an FMO input file

fragit --use-config=myconfig.in F709_reference.pdb.txt
F709_reference.pdb.txt - this is a .pdb file

which successfully yield a GAMESS-FMO input file.
F709_reference.inp.txt - this is a .inp file

Running this calculation leads to the computation of all fragments against all fragments 128x128 (127 amino acids + 1 ligand) which is undesirable. The calculation however runs (seemingly) fine.

However, when I change in the myconfig.in
centralfragment = 0 to centralfragment = 128
and set
--output-active-distance to --output-active-distance 6

it returns a file that crashes the calculation.

Could you, please, suggest a way to obtain the desired fragmentation scheme?

Additional question: I would also be interested in further "fragmenting" the fragments between side chains and backbone to calculate their individual contributions. Would this be possible with your tool and if so, I wold be grateful if you could indicate me how.

Many thanks in advance for your response,

Bests,

Harold Grosjean

[cstein]

Hi Harold and thanks for opening an issue,

Running this calculation leads to the computation of all fragments against all fragments 128x128
(127 amino acids + 1 ligand) which is undesirable. The calculation however runs (seemingly) fine.

If I understand you correctly, you are only interested in printing out information for a ligand and some of the interacting residues. For this you still need to run FMO where all interaction between all (128) fragments are taken care of. I know there are a lot of information being printed but you need to write a small script to extract the information you want.

However, when I change in the myconfig.in
centralfragment = 0 to centralfragment = 128
and set --output-active-distance to --output-active-distance 6
These options are specifically for running geometry-optimizations with the FMO/FD method (https://pubs.acs.org/doi/pdf/10.1021/jz1016894). I will try and install GAMESS and see why this crashes based on your structure, but maybe options have changed over the years or some other mysterious things happen. I will look into it. Thanks!

Additional question: I would also be interested in further "fragmenting" the fragments between side chains and backbone
to calculate their individual contributions. Would this be possible with your tool and if so, I wold be grateful if you could
indicate me how.
Fragmentation works by specifying pairs of atoms to fragment. The power of fragit comes from using SMARTS (https://www.daylight.com/dayhtml/doc/theory/theory.smarts.html) to specify this. For proteins you can see the way fragmentation is carried out in the [fragmentpatterns] group of the config file. It looks like this

[fragmentpatterns]
peptide = [$(CN)][$(C(=O)NCC(=O))]

The code basically looks for a carbon atom on one side of a bond connected to a nitrogen $(CN), and it looks for a carbon atom connected to a doubly bound oxygen atom and a nitrogen as well as a little more, i.e. $(C(=O)N .... If you want to fragment the backbone from the side chains you would need to insert additional lines to fragment specific bonds.

Don't hesitate to write

Casper

[haroldgrosjean]

Hi Casper,

So what I am thinking of doing is extract the binding site in a separate pdb and add hydrogens to the "artificial" ends so that the hybridization of these atoms remains correct during FMO calculations.

So I did that with MDAnalysia and rdkit which yielded this file:
F126_BS_H.pdb.txt

If I run fragit on this pdb file alone with standard configuration and almost got the desired results but it seems to include/ exclude protons from neighboring fragments.

Then, I tried to add: mod_ct =[H][C](-[N])-[C]([H])=O in the protection pattern section to protect the artificial C-term it further confused the software and the fragmentation gets worst.

I am not in expert is this particular area of chemioinformatics and would very much appreciate if you could help me define an appropriate fragmentation/ protection scheme.

Many thanks in advance,

Harold