[buster-discuss] rhofit, use maps instead mtz, or specify map coefficients?

[Guenter Fritz]

Dear Clemens and Andrew,

thanks a lot, that answered most aspects of my questions!

Here some (I hope not too much)  info what I am doing:

I am comparing currently several tools for automated ligand fitting into 
maps. The background is simple: We did some fragment screening campaigns 
and we were tired of looking into more than 1000 datasets.
So I started to explore the different tools and implemented several in a 
pipeline. I am aware that there is already such an excellent pipeline 
-pipedream :).

Nevertheless, not everybody is using GPhl software and I compared how 
the different tools perform in different settings.
I had a look:
What is effect, if you use different refinement programs in the first 
step to create maps .
Which program is fastest and can place the ligand most accurately.
How to deal with side chain movements, water placement and subsequent 
refinement .
And finally how to evaluate the fitted fragment in an automated way to 
get some info how good is the fit, where does it bind, and what might be 
the affinity.

I did some bash scripting; I can feed the pipeline now with an excel 
list of my datasets and ligands (or different pdbs) and get refined 
complexes. As summary an excel list with all relevant info is created as 
well as  a html with some figures of the fitted ligands in the 
densities. I think it has become a nice tool.

Now I want to publish it and for that I want to do some more comparisons.
When I use maps (mtz files) from buster (using -L option) or  refmac, 
flynn and rhofit perform both very good. Now I wanted to check how both 
programs compare fitting into event maps from pandda and check how my 
basic script performs for published datasets even with some low 
occupancy fragments.

flynn from Open Eye can use directly the event maps (ccp4 map files) 
from pandda.
Therefore I wanted to know, whether this is possible for rhofit as well.

As a workaround I have transformed the event maps with cinvfft into mtz 
files (result is OK) for rhofit and wanted to know how to define the map 
coefficients. And thanks for the suggestion, yes I could simply renmae 
those with mtzutils.

Best regards,
Guenter
> Dear Guenter,
>
> On Mon, Oct 17, 2022 at 09:17:59AM +0200, Guenter Fritz wrote:
>> I was wondering whether one could use directly maps for rhofit or whether it
>> is possible to specify the map coefficients of the mtz to be used by rhofit.
> It is possible in a maybe slightly round-about way ...
>
> Rhofit can also start from a set of already defined so-called
> "clusters" - which would be the result of internally running
> "prep_rhofit" on the given MTZ file and some default map
> coefficients. What you need is a simple text file with a list of
> cluster PDB files (one per line, i.e. several lines if you have
> several binding sites you want Rhofit to analyse). Each cluster PDB
> file contains a set of grid points with an associated map score
> (usually: mFo-DFc map value) that is generated e.g. by prep_rhofit,
> which has more control about the coefficients to use for map
> calculation (see "prep_rhofit -h" for details).
>
> Of course, you could always rename the map coefficients of your input
> MTZ file e.g. via
>
>    mtzutils hklin my.mtz hklout tmp.mtz <<e
>    COLU MY_F=2FOFCWT MY_P=PH2FOFCWT
>    e
>
> If you have only one set of map coefficients: you need to give them as
> both (2mFo-DFc and mFo-DFc) to prep_rhofit.
>
> In principle one can prepare such cluster files also directly from
> maps: if you could give a bit more background information maybe we can
> provide further information for you here?
>
> Cheers
>
> Clemens & Andrew