Call CrystalNets from Python?

[qzhu2017]

Hi, this is really a nice package. I have some python script that would generate many cif files. I want to use this package to determine the topology and then pass the value to my python script.

Do you know if I can call CrystalNets from Python?

Many thanks,

[Liozou]

Hi and thank you for your interest!

Sure, you can call julia from python using juliacall. To do so, go to the folder with your python script and add a new file called juliapkg.json with the following content:

{
    "julia": "1",
    "packages": {
        "CrystalNets": {
            "uuid": "7952bbbe-a946-4118-bea0-081a0932faa9",
            "version": "0.3"
        }
    }
}

then install juliacall with pip install juliacall. From there, you can call CrystalNets by starting your python script with the following three lines:

import juliacall
jl = juliacall.newmodule("TheNameOfYourModule") # put whatever name here
jl.seval("using CrystalNets")

(Note that the first time you run those, julia will install CrystalNets and its dependencies, which can take a while, but you only need to do that once)

Afterwards, you can call any function of CrystalNets.jl by prefixing them with jl., for instance:

>>> jl.determine_topology("/tmp/MIL-53.cif")
[ Error: Atom ?, used in a bond, has either zero or multiple placements in the CIF file. This invalidates all bonds from the file, which will thus be discarded.
[ Warning: Guessing bonds with custom algorithm (from Chemfiles and VMD). This may take a while for big structures and may be inexact.
[ Info: To avoid guessing bonds, use a file format that contains the bonds.
Export of input is enabled: saving file at /tmp/input_MIL-53_other1.vtf
Export of subnet_Auto is enabled: saving file at /tmp/subnet_Auto_MIL-53_1_other1.vtf
Julia: rna

(rna is the name of the topology of MIL-53)

If you want, you can disable the warnings and/or the exports by adding the following lines after the three initial lines above:

jl.CrystalNets.toggle_warning(False) # to disable warnings
jl.CrystalNets.toggle_export(False) # to disable exports

A more programmatic use-case would then look like

>>> x = jl.determine_topology("/tmp/UiO-66.cif", jl.CrystalNets.Options(structure=jl.StructureType.MOF))
>>> str(x)
'AllNodes, SingleNodes: fcu'

The nature of the x above depends on whether your cif files contains inter-penetrating substructures, or which clusterings you asked, there are examples for all cases in the documentation.

Please let me know if you encounter any other issue with this!

[Liozou]

I included the explanation above, as well as a slightly more detailed example, in a new Python interface tutorial now accessible in the documentation: https://molsim.info/CrystalNets.jl/dev/python Let me know if you'd like to see something else in the tutorial!

[qzhu2017]

Hi, @Liozou Can you tell me how to extract the dimension information if I choose the auto option as follows

result = jl.determine_topology(cif_file, jl.CrystalNets.Options(structure=jl.StructureType.Auto))

[Liozou]

Sure, the code at https://molsim.info/CrystalNets.jl/dev/python/ should contain the necessary information. Note that specifying structure=jl.StructureType.Auto is not necessary since this is the default value for the structure keyword anyway (check https://molsim.info/CrystalNets.jl/dev/lib/public/#CrystalNets.Options).

In general the dimension depends on the net, and thus it may depend on the clustering. So, to detail the steps in the current version of CrystalNets, (v0.4), you should: 1) Extract the relevant TopologyResult, so if there are multiple interpenetrated substructures and you want the second one for instance, you can do tresult, tresultnfold = result[1] where tresultnfold will be the catenation multiplicity of the net. If you want the first one, do tresult, tresultnfold = result[0], etc. 2) Extract the TopologicalGenome of the net with the relevant clustering, so if you are interested in the SingleNodes clustering for instance, you can do genome = tresult[jl.Clustering.SingleNodes]. Note that, if you have not specified explicitly the clusterings keyword, then:

• If your structure is not StructureType.MOF (like in your example, where it is StructureType.Auto, which is also the default), then the default clusterings is [Clustering.Auto] so you should query genome = tresult[jl.Clustering.Auto]
• If your structure is StructureType.MOF then the default clusterings is [Clustering.AllNodes, Clustering.SingleNodes] (see https://molsim.info/CrystalNets.jl/dev/lib/public/#CrystalNets.StructureType), so you should query genomeSingleNodes = tresult[jl.Clustering.SingleNodes] as well as genomeAllNodes = tresult[jl.Clustering.AllNodes] and probably handle the case where they differ somehow. 3) Obtain the dimension like so: dim = jl.ndims(jl.CrystalNets.PeriodicGraph(genome))