Machine Learning-Driven Multiscale Modeling: Bridging the Scales with a Next-Generation Simulation Infrastructure

Abstract

Interdependence across time and length scales is common in biology, where atomic interactions can impact largerscale phenomenon. Such dependence is especially true for a wellknown cancer signaling pathway, where the membrane-bound RAS protein binds an effector protein called RAF. To capture the driving forces that bring RAS and RAF (represented as two domains, RBD and CRD) together on the plasma membrane, simulations with the ability to calculate atomic detail while having long time and large length-scales are needed. The Multiscale Machine-Learned Modeling Infrastructure (MuMMI) is able to resolve RAS/RAF protein−membrane interactions that identify specific lipid−protein fingerprints that enhance protein orientations viable for effector binding. MuMMI is a fully automated, ensemble-based multiscale approach connecting three resolution scales: (1) the coarsest scale is a continuum model able to simulate milliseconds of time for a 1 μm 2 membrane, (2) the middle scale is a coarse-grained (CG) Martini bead model to explore protein− lipid interactions, and (3) the finest scale is an all-atom (AA) model capturing specific interactions between lipids and proteins. MuMMI dynamically couples adjacent scales in a pairwise manner using machine learning (ML). The dynamic coupling allows for better sampling of the refined scale from the adjacent coarse scale (forward) and on-the-fly feedback to improve the fidelity of the coarser scale from the adjacent refined scale (backward). MuMMI operates efficiently at any scale, from a few compute nodes to the largest supercomputers in the world, and is generalizable to simulate different systems. As computing resources continue to increase and multiscale methods continue to advance, fully automated multiscale simulations (like MuMMI) will be commonly used to address complex science questions.

Membrane lipids drive formation of KRAS4b-RAF1 RBDCRD nanoclusters on the membrane

“…The MuMMI infrastructure was initially developed to bridge two-scales: a large and long timescale macro model simulation with an ensemble to selected higher resolution 56 coarse-grained (CG) MD simulations 18 , 57 . Recently, MuMMI was greatly extended to integrate three-scales: the macro scale and CG capabilities were broadened to include All-Atom (AA) simulations to capture the atomistic details of significant events 56 . The new MuMMI was used to simulate RAS and RAS-RBDCRD dynamics on the PM 21 .…”

“…The new MuMMI was used to simulate RAS and RAS-RBDCRD dynamics on the PM 21 . The main improvements to MuMMI are: support and parameters for an additional protein (RAS-RBDCRD) 26 , 56 , a new AA with reliable CG-to-AA transformations 58 to sample changes in protein secondary structure, a new machine learning sampling framework used to select simulations of interest at finer resolutions (macro to CG and CG to AA) 19 , a new faster and higher fidelity macro model 59 , and an updated workflow that is generalizable and has extend scalability and fault tolerance 60 .…”

“…The MuMMI RAS-RBDCRD simulation campaign 56 consisted of a large macro model continuum simulation (1000 nm X 1000 nm) with ~150 KRAS and ~150 KRAS-RBDCRD proteins. The macro model samples realistic fluctuations in the local composition of the PM, KRAS and KRAS-RBDCRD colocalization, their lipid preferences, and changes in protein orientation relative to the membrane (the protein orientational state).…”

MLatom 3: A Platform for Machine Learning-Enhanced Computational Chemistry Simulations and Workflows

“…Computational chemistry simulations are common in chemistry research thanks to abundant general-purpose software, most of which have started as purely quantum mechanical (QM) and molecular mechanical (MM) packages. More recently, the rise of artificial intelligence (AI)/machine learning (ML) applications for chemical simulations has caused the proliferation of programs mostly focusing on specific ML tasks such as learning potential energy surfaces (PESs). − The rift between the development of the traditional QM and MM packages on the one hand and ML programs on the other hand is bridged to some extent by the higher-level library ASE, which enables usual computational tasks via interfacing heterogeneous software. The further integration of QM, MM, and ML has been prompted by the maturing of ML techniques and is evidenced by the growing trend of incorporating ML methods in the QM and MM computational chemistry software. ,− …”

Mesoscopic Model for Reversible Adsorption Stage of Albumin and Fibrinogen on TiO2 Surface

“…Up to now, existing coarse-grained models include MARTINI, MS-CG, UNRES, OPEP, PRIMO, SIRAH, REM,and so forth. Moreover, machine learning was also used to analyze full-length MD trajectories and to build a multiscale model …”