NMRlipids20 workshop outcomes

Thirteen contributors participated to the second NMRlipids workshop organized on 6th-9th of September in Prague, Czech Republic. 

Three talks on topics related to the NMRlipids project were given on the first day: Pavel Jungwirth presented results on how charge scaling can improve MD simulation quality without additional computational cost, Hanne Antila presented results from a automatic force field optimization algorithm, and Ricky Nencini presented evaluation of drug molecule binding affinities in MD simulations against NMR data.

Batuhan Kav presented the current status of the NMRlipids VI project on polarizable force fields. The main conclusion from the talk and consequent discussion was that the headgroup conformational ensemble and ion binding in Drude model is significantly worse than in original CHARMM36, most likely because forking and signs of headgroup order parameters are not taken into account during parameter optimization. The results from AMOEBA simulations are not yet available due to practical issues, but consensus was that those would be highly relevant.

Second day was opened by Samuli Ollila with the presentation on the current status of the NMRlipids databank. This was followed by work in three groups:

1. Quality Evaluation. Goal: Define the quality measures for order parameters, find robust code for form factor calculation and include this into the quality measure. Outcomes: Quality measure for order parameters, S=-log(P), where P is the probability mass within the experimental error for a normal distribution around simulation result (mean = order parameter from simulation, standard deviation = error of order parameter from simulation). Implementation of this was preliminary tested and committed to the QualityEvaluation.py. A new form factor code with python is now implemented, preliminary tested, and committed into the MATCH repository. 
2. Extension of the databank to other than Gromacs programs. Goal: Define the required files needed to add simulations ran with other than Gromacs program into the databank. Outcomes: Possibility to incorporate OpenMM data was implemented. The required files are the trajectory (e.g., in dcd format), structure (e.g., in pdb format) and either xml file or inp file. Force field / topology information can be optionally given as psf file. This is implemented in branch of Anne Kiirikki, but not yet merged to main branch.
3. Analysis of the data. Goal: Test and improve the codes for the analysis of the databank. Outcomes: A new class that makes the analysis of databank significantly easier was introduced. An example code was implemented and commited. Also a new way to organize the information about molecular composition in the databank README.yaml files was proposed to ease the writing of analysis codes. The implementation of this is currently in progress by Anne Kiirikki.

In addition to these, several other topics were touched during discussions. These include, for example, practical ways to handle united atom simulations, stereo specific information in mapping files, sanity checks for the data, NMRlipids III project, and other potential data repositories than Zenodo. However, practical steps to progress these topics were not taken.

In conclusion, the workshop was again highly useful, at least for the NMRlipids project, and hopefully also for the participants. Thank you for all the participants!