ML-driven multiscale modeling - next gen simulation infrastructure

Data for Machine Learning-driven Multiscale Modeling: Bridging the Scales with a Next Generation Simulation Infrastructure

Ingólfsson H.I., H. Bhatia, F. Aydin, T. Oppelstrup, C.A. López, L.G. Stanton, T.S. Carpenter, S. Wong, C. Neale, F. Di Natale, X. Zhang, J.Y. Moon, C.B. Stanley, J.R. Chavez, K. Nguyen, G. Dharuman, V. Burns, R. Shrestha, D. Goswami, G. Gulten, Q.N. Van, A. Ramanathan, B. Van Essen, N.W. Hengartner, A.G. Stephen, T. Turbyville, P.-T. Bremer, S. Gnanakaran, J.N. Glosli, F.C. Lightstone, D.V. Nissley, F.H. Streitz. 2022. Machine Learning-driven Multiscale Modeling, Bridging the Scales with a Next Generation Simulation Infrastructure. Journal of Chemical Theory and Computation. Accepted. doi: 10.1021/acs.jctc.2c01018. 

The Multiscale Machine-Learned Modeling Infrastructure (MuMMI) was expanded to include three resolution scales: 1) a continuum macro model able to simulate milliseconds of time for a micron-squared membrane, 2) a coarse-grained (CG) Martini model used to explore protein-lipid interactions, and 3) an all-atom (AA) CHARMM36m model capturing specific interactions between lipids and proteins. The updated three-scale MuMMI was used to simulate RAS and RAS-RAF (RBD and CRD domains) dynamics on a plasma membrane.

The MuMMI updates and results of simulation campaign are described in several additional publications, including:

  • Stanton, L.G, T. Oppelstrup, T.S. Carpenter, H.I. Ingólfsson, M.P. Surh, F.C. Lightstone, J.N. Glosli. 2023 Dynamic Density Functional Theory of Multicomponent Cellular Membranes. Physical Review Research. 5:013080, doi: 10.1103/PhysRevResearch.5.013080.
  • Bhatia H., J.J. Thiagarajan, R. Anirudh, T.S. Jayram, T. Oppestrup, H.I. Ingólfsson, F.C. Lightstone, P.-T. Bremer 2022. A Biology-Informed Similarity Metric for Simulated Patches of Human Cell Membrane. Machine Learning: Science and Technology. 3:035010, doi: 10.1088/2632-2153/ac8523.
  • Nguyen K, C.A. Lopez, Q.N. Van, C. Neale, T.S. Carpenter, F. Di Natale, A. Chan, T.H. Tran, T. Travers, H. Bhatia, X. Zhang, T. Reddy, V. Burns, J.N. Glosli, T. Oppelstrup, N.W. Hengartner, P.-T. Bremer, D. Chen, R. Shrestha, C. Agamasu, T. Turbyville, D.K. Simanshu, F.H. Streitz, D.V. Nissley, H.I. Ingólfsson, A.G. Stephen, F.C. Lightstone, S. Gnanakaran. 2022. Exploring CRD mobility during RAS/RAF engagement at the membrane. Biophysical Journal, 121:3630-3650, doi: 10.1016/j.bpj.2022.06.035.
  • Lopez C.A., X. Zhang, F. Ayden, R. Shrestha, Que N. Van, C.B. Stanley, T.S. Carpenter, K. Nguyen, L.A. Patel, D. Chen, V. Burns, N.W. Hengartner, T. Reddy, H. Bhatia, F. Di Natale, T.H. Tran, A.H. Chan, D.K. Simanshu, D.V. Nissley, F.H. Streitz, A.G. Stephen, T.J. Turbyville, F.C. Lightstone, S. Gnanakaran, H.I. Ingólfsson, C. Neale. 2022. Asynchronous Reciprocal Coupling of Martini 2.2 Coarse-Grained and CHARMM36 All-Atom Simulations in an Automated Multiscale Framework. Journal of Chemical Theory and Computation, 18:5025-5045, doi: 10.1021/acs.jctc.2c00168.

Here we provide input parameters and examples used at the different scales.

The raw simulation data ~270 TB is accessible on the NIH MoDaC server:

MuMMI is composed of  number of sub-components -- both freely-available third-party applications and custom codes -- including:

This work was performed under the auspices of the U.S. Department of Energy (DOE) by Lawrence Livermore National Laboratory (LLNL) under Contract DE-AC52-07NA27344, Los Alamos National Laboratory (LANL) under Contract DE-AC5206NA25396, Oak Ridge National Laboratory under Contract DE-AC05-00OR22725, Argonne National Laboratory (ANL) under Contract DE-AC02-06-CH11357, and under the auspices of the National Cancer Institute (NCI) by Frederick National Laboratory for Cancer Research (FNLCR) under Contract 75N91019D00024. This work has been supported by the Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) program established by the U.S. DOE and the NCI of the National Institutes of Health. LLNL-MI-847086


All data files (40 MB): RAS_RAF_parameter_data.tar.gz

Macro-scale model input files

All macro-model parameters are provided in macro-c3-input-files.tgz. This includes lipid-lipid direct correlations functions, protein-lipid potentials (for both RAS and RAS-RAF each in three different orientational states), protein state transition rates, diffusion rates, etc., as well as initial input files and positions.

Coarse-grained simulation input files

Martini 2 ( was used for the coarse-grained simulations and most of the input parameters are the same as those used in the RAS-only two-scale MuMMI campaign and can be accessed here: RAS-lipid-dependent-dynamics-data.html

The parameters that were updated include PIP2 with a pH adjusted charge (now -4), martini_v2.0_PAP6_02.itp, and updated protein topologies for KRAS4b, RAS_scfix-CYFpos.itp, and KRAS-CRAF (RBD and CRD domains), RAS_RAF_TERNARY_scfix-CYFpos.itp. This RAS-RAF topology is with feedback (changes in secondary structure) derived from the AA simulations in the RAS-RAF multiscale simulation campaign (same as parameters in final CG simulation run but different from the initial simulations). These protein parameters are based on an initial version from LANL which can be found here:

All-atom simulation input files

The AA CHARMM36 input parameter files can be accessed with the backmapping files here:

For convenience, we have provided an extracted collection of the production topology files (aa-params.tar.gz).

Sample RAS-RAF CG simulation files (pfpatch_000000145718) - GROMACS topology file containing the list of all particles in this simulation.

initial-system-CG.gro - Initial GROMACS particle coordinate file for this simulation (after system setup and initial equilibrium)

Sample RAS-RAF AA simulation files (pfpatch_000000145718_f000202000000) - GROMACS topology file containing the list of all particles in this simulation.

backmapped-system-AA.gro - GROMACS particle coordinate file for this simulation (after initial system backmapping and initial equilibrium)