Public Research Seminar by Advanced Materials (AMAT) Thrust, HKUST(GZ) - Evaluating and developing molecular dynamics models in the era of open-access, large-scale data

Classical molecular dynamics simulations (MD) are a widely used tool in many fields of science ranging from biophysics to material science. However, the accuracy of molecular dynamics simulations hinges on the ability of force fields---the models used in MD simulations to describe the interactions of particles in the system---to represent the reality. Hence, the initial step to any work using MD simulations should be ensuring at the chosen model correctly captures the aspects of the simulated system that are relevant for the research problem at hand. Failure to do so can compromise the quality of the results as many of the current models have severe flaws, such as producing wrong dynamics or conformations for the system [1].

The emerging era of open-access, large-scale data brings forth new tools for both bechmarking and and further developing force fields [2]. In the case of biomembrane systems (lipid bilayers), the NMRLipids open science project [3] has for the last decade spearheaded the critical evaluation of current force fields and development of open data practises to do so. Recently, this culminated in the publication of the NMRlipids databank [4], which enables (large scale) simulation quality evaluation, provides data for machine learning applications, and has already enabled user cases beyond the reach of regular MD simulation volumes available for single research group. In many sense, lipid bilayers are an ideal test case for these efforts. As the core of the cell membrane, lipid bilayers are highly biologically relevant and they play role in the cutting edge drug delivery applications. Good amount of high resolution experimental data is available, and typical bilayer simulations consist of several lipid molecules which promotes good sampling for the comparison to experiments. That said, all tools and practises presented here are transferable to other systems.
----------------------------------
References: 
[1] H. S. Antila, S. Dixit, B. Kav, J. J. Madsen, M. S. Miettinen, O. H. S. Ollila, Evaluating
polarizable biomembrane simulations against experiments. J. Chem. Theory Comput,
20 (10), 4325–4337, 2024
[2] H. S. Antila, B. Kav, M. S. Miettinen, H. Martinez-Seara, P. Jungwirth, O. H. S. Ollila
Emerging Era of Biomolecular Membrane Simulations: Automated Physically-Justified Force
Field Development and Quality-Evaluated Databanks, J. Phys. Chem. B, 126 (23),
4169–4183, 2022
[3] nmrlipids.blogspot.com
[4] A. M. Kiirrikki, H. S. Antila, L. S. Bort, B. Buslaev, F. Favela-Rosales, T. Mendes Ferreira,
P. F. J. Fuchs, R. Garcia-Fandino, I. Guschin, B. Kav, N. Kucherka, P. Kula, M. Kurki,
A. Kuzmin, A. Lalitha, F. Lolicato, J. J. Madsen, M. S. Miettinen, C. Mingham, L.
Monticalli, R. Nencini, A. M. Nesterenko, T. J. Piggot, A. Pineiro,N. Reuter, S. Samantray,
F. Suarez-Leston, R. Talandashti, O. H. S. Ollila, Overlay databank unlocks data-driven
analyses of biomolecules for all. Nature Comm. 15, 1136, 2024.