openforcefield / openff-qcsubmit

Automated tools for submitting molecules to QCFractal
https://openff-qcsubmit.readthedocs.io/en/latest/index.html
MIT License
26 stars 4 forks source link

Standard scan range for improper/non-rotatable torsions #20

Open jthorton opened 4 years ago

jthorton commented 4 years ago

ISSUE we have the ability to scan improper/ring-torsions now, but should we automatically apply some default limit on the scan range? or will torsiondrive cope with standard settings and stop grid points over a certain energy threshold?

TODO

jchodera commented 4 years ago

I'd be surprised if the standard settings worked, but it's worth a try! A limit in torsional displacement would be simple to implement, but if it was straightforward to develop, perturbations up to a limit in strain energy (above the minimum) would seem to be a more useful parameter to specify.

davidlmobley commented 4 years ago

I believe it stops grid points over a certain threshold, but that'd be a question for @leeping . I am fairly certain he has this in place though. Still, some default limit on the scan range may make sense.

@jmaat also may have insights as this has some similarity to things she was thinking about in her work on impropers. Or at least she may also know the answer.

leeping commented 4 years ago

TorsionDrive has the option to specify limited ranges of angles and an upper limit to the energy, but they don't come by default and you need to set them. I think this should be available through both the command line and API but @yudongqiu would know more details.

jmaat commented 4 years ago

From my work, I found that scanning from -50 to +50 degrees around the improper angle would typically capture the full minima. This the the range we used for the improper grid optimization datasets in QCA. I would suggest a limit from -60 to 60 for the improper angle based off of these results.

jthorton commented 4 years ago

Thanks for the advice it seems like Torsiondrive is naturally handling this problem with its energy cut off, I have found a few ring torsions in the archive already which appropriately limit the scan range of ring torsions even when the full scan range is requested. I have also introduced a way for users to specify a different scan range if they wish to limit them to the range @jmaat suggests though.