Fragment Positioning
The idea behind the positional run is, to make an stochastic run on several
points of the surfaces of an active sites (alias positions) with an
interesting small molecule (alias functional group) in "all" orientations.
To do a positional run, you need to go through the following steps:
- create the position with the target enzyme (or molecule)
- cut out the active site
- create all functional groups you want to test
- set up the parameters for the positional run
- start a job for each functional group
Remark: Step 1 - 4 can be done in any order.
Discussion of each step:
1 create the position with the target enzyme (or molecule)
->Top
For preparation, set a small molecule into the active site of the enzyme.
Be sure to use the whole enzyme, otherwise the program gets irritated by fake
surfaces.
Use cnf/g/c to create the positions. Select the enzyme. You are now in the set
menu. You have to form a set of all atoms of the surface of the active site.
The easiest way to do it (in my opinion):
- select the small molecule in the active site (option e)
- select the surface of the active site (option d; answer the question
"entire monomers?" with no)
- correct it by visual inspection (option a)
- leave the set menu (option x)
Now you will see the (purple dotted) surface of the active site.
The program prompts you for the minimal # of points. It means: How many dots
of the surface needs an atom at least to be considered as a surface atom?
The whole surface has 30 points, so 4 points means that more than a eighth
of the surface is exposed, which is clearly a surface atom. For each surface
atom there will be a positions. Recommendation: take a number between 1 on 4.
The higher this number the less positions you get.
Now the positions are displayed.
The program prompts for the minimal distance between two positions. If two
positions are closer than this distance, they will be replaced by one position
in the middle of them. A distance of about 1 A is recommended. default is 0.
The last prompt will ask you for a name for these positions and gives you also
the number of points. You will return to the cnf/g menu.
The cnf/g/d option lets you delete single points of these positions.
The cnf/g/r option removes the points at the surface. The algorithm uses, that
the active site is a cavity. This option can only be used if the active site
is concave. For details of the algorithm see release notes of release 2.5.
Positions can be stored in .jkl files by (.../s/j) and restored by (.../g/j).
2 cut out the active site
->Top
That is the same as with the stochastic run. The most important points are:
- The smaller the active site the faster the run
- All atoms that come in contact with the ligand should be included
- Don't change the chemical behavior of these atoms (i.e. don't cut residues)
3 create all functional groups you want to test
->Top
Just make your favorite functional group. If possible, they should not have
any conformational freedom.
It is recommended to store them in a CIF file, so you have them handy anytime.
4 set up the parameters for the positional run
->Top
Whether a run gives the results you wish or not, whether a run is fast or not,
that depends mainly on the setting of the parameters. So this is really the
most important part.
Once you set your parameter, you can store them on the
file system (cnf/g/p/s) and read them again (cnf/g/p/g).
btw: If you run in batch mode, the parameters are stored in a file.
In the file positional.Fpar the recommended parameters are stored. So you can
just read that file (cnf/g/p/g) or you set the parameters with the cnf/g/p/p
option. If you read the file, you still should change the parameter "job name".
Here the important parameters for a positional run:
mode: must be set to positional.
job name: _%s is recommended. The %s stands for
the name of the functional group. You need that in the job name to avoid
confusion between the different runs.
job control: batch is recommended.
input source: entry. Otherwise the orientation library is
not generated.
mode of random trials: generated structure is recommended.
# random trails: 1. Since the functional group has no
conformational freedom, you do not want do explore the conformational space.
Flexible functional groups are not recommended. Higher numbers will slow down
the job without giving better results.
rmsd tolerance: 1.0 is recommended. You only want to do a
rough survey, where 1.0 A is sufficient. This parameter influence the
performance of the job very strongly:
- The start orientations differ at least by this amount in rmsd.
Though: the larger this tolerance, the less orientations need to be tested.
This will be done automatically.
- The accuracy of the minimization can be adjusted and will speed up a
single minimization. This parameter needs to be adjusted manually (see below).
accuracy postrelaxation: about 1/10 of rmsd tolerance. If it is larger,
the same position may be considered twice only because the minimization
accuracy was not suf ficient. If it is smaller, a lot of computation time is
wasted in accuracy that is not needed for this job.
Be sure that the "accuracy relaxation" is bigger than the "accuracy
postrelaxation", otherwise enlarge it. The same with "accuracy
preminimization" & "accuracy relaxation". If you fail to do that, the bottle
neck will be in the relaxation or preminimization, respectively.
The other parameters are not so important for positional run. The default
values are recommended. Do not change any parameter if you don't have a
reason for it.
5 start a job for each functional group
->Top
Use the cnf/g/m. Select all functional groups. Select your active site.
In the positional run menu select the positions you have created in step (1)
(option s). Select the queue you want the job submit to.
You will get a conformation library for each of the functional group.
The new multi-library display (cnf/m) will help you to visualize the results.
Because all conformation in a library are the same (the differ only by
position) be sure to calculate rmsd always without superposition!