open force field

An open and collaborative approach to better force fields

Oct 16, 2019 Advisory Board Meeting

Posted on 16 Oct 2019 by Karmen Condic-Jurkic

The Open Force Field Consortium Advisory Board had an in-person meeting during the Consortium workshop in San Diego (Aug 30-31, 2091), discussing research questions of interest and potential future directions for the Open Force Field Initiative. This and other discussions held during the meeting made a basis for the roadmap for Year 2. This was the joint August/September meeting.

The minutes from the meeting held on Oct 16, 2019 are summarized as follows:

Parsley release

  • The first optimized Open Force Field 1.0 has been released under the codename “Parsley” and there are 2 different versions of this force field – one parameterized with H-bond constraints and one without. Differences between these two versions are minor. More details about the entire optimization process can be found in the blog post. This force field has been fit primarily to QM data and served as a test for parameterization infrastructure developed. The optimized force field showed general improvement in optimized geometries (80 %) and bonding parameters (Pfizer “discrepancy” test and Zinc FDA drugs used for benchmarking). Initial geometries were generated with OE Omega.
  • Suggestions for future benchmarking:
    • Compare RMSD and energies for better agreement
    • Compute RMSF with respect to previous geometries and non-equilibrium geometries
    • Comparison with other FFs and ANI
    • Comparison with data used to train ANI
  • Charge support – at the moment, OpenFF is using AmberTools or OpenEye BCC implementation, but a more systematic approach to charge assignment and adequate support is a work in progress.
  • Comparison with OPLS3 is highly desirable, but may not be easily achieved due to limitations imposed by Schrodinger license. The question how to include OPLS3 in benchmarking remains unresolved at the moment, but there are some options to explore.
  • Coverage tool – pharma partners reported high coverage for their internal molecule sets (> 99%).
  • A tool for parameter quality check would be desirable, but unlikely to be developed in the next 6 months due to limited resources. T. Gokey (Mobley group) will do some work to determine whether it’s possible to quantify “implied” errors in parameterization. This is one of the research questions in the roadmap for Year 2.

Roadmap

  • OpenFF has created a roadmap for Year 2 based on discussions held during the in-person meeting in August in San Diego. It’s a series of research questions that will be explored in Year 2 and supported with necessary software development.
  • Interest in bespoke pipeline has been expressed (for specific molecules) and Josh Horton will likely take over this line of OpenFF development. Josh is a newly hired postdoc for OpenFF and he will start in November in Lancaster after graduating in Newcastle. He worked on QUBEKit during his PhD.
  • The current tools for FF parameterization are available, but they are not super user-friendly at the moment. There are plans to improve documentation and provide more tutorials and instructions on how to use the existing tools.

CCG support for SMIRNOFF format

  • D. Mobley probed if there is interest to ask for support for SMIRNOFF format by CCG and what would be the best way to implement it. R. Walker offered to look into what it would take to include an external FF API-wise with his student and also suggested that having ParmEd to convert beween FFs would be a desirable feature. P. in’t Veld gave pointers to their internal tool (montecarlo.sourceforge.net) which handles a number of FFs, including COMPASS, CHARMM, OPLS, etc and they are currently building an AMBER plugin to load SMIRNOFF. Request for SMIRNOFF support should come from pharma partners.

Prorated membership

  • OpenFF team suggested allowing prorated membership for new members joining later in the funding year, which starts on Oct 1. There is a general support for this, but this change in RCA could potentially be included only for Year 3 for legal reasons. This is to be checked with pharma legal departments.

Logo slide

  • OpenFF had to put in trademark permission request for several members to allow use of their logo for the Consortium. Still waiting for approval from everyone to use logo slide.

Other topics

  • OpenFF is working on getting federal funding.
  • OpenEye is formally joining the Consortium.
  • The next in-person meeting could potentially take place in Boston before or after the Alchemical Free Energy meeting in May, 2020.