Implementation of Girsanov reweighting in openmmtools integrator module

[anyschaefer]

Description

We extended the existing Langevin integrator class, such that the reweighting factors are calculated on-the-fly and, at regular intervals, are written to a reweighting factor trajectory file.

Status

• Ready to go

Reference

Joana-Lysiane Schäfer, Bettina G. Keller, Implementation of Girsanov reweighting in OpenMM and Deeptime

[anyschaefer]

Langevin integrator with Girsanov reweighting

[sukritsingh]

Awesome! Would this arxiv paper be the the relevant reference for this reweighting procedure? https://arxiv.org/abs/2303.14696

[ijpulidos]

This is a great contribution. Thanks a lot for your interest in using our tools and extend them with new features.

@anyschaefer would you be able to provide a few minimal scripts or examples that we could use to write tests for this new integrator? We would very much appreciate your input on what actually should be tested, or how are you validating it at your lab.

@sukritsingh thanks for the references, that should also help when validating/testing this new feature.

Thanks again, this is looking great!

[anyschaefer]

@sukritsingh Exactly, this is the theoretical basis for the underdampt Langevin integrator schemes with Girsanov reweighting. The reference for the implementation in openmmtools is currently under review and will soon be publicly available. I will then link it here immediately.

@ijpulidos We have tested the new modules on analytical and various molecular models, such as the I-Ca-I dissociation process in water. The corresponding discussion and a detailed description of the implementation will soon be available with "Implementation of Girsanov reweighting in OpenMM and Deeptime".

With the repository reweightingtools we provide all input files as well as the setup for corresponding simulations and their evaluation.

[anyschaefer]

I am pleased to provide you with a link to the publication Implementation of Girsanov Reweighting in OpenMM and Deeptime.

Would you also be interested in adding the corresponding reweighting reporter function to openmmtools?

[ijpulidos]

Thanks for the references, these are very helpful. I asked because we want to have tests for this if we want to merge these changes (both unit and integration tests).