Machine Learning-Driven Multiscale Modeling: Bridging the Scales with a Next-Generation Simulation Infrastructure
Not Available
Engineering topics
Publications and source records attributed to Di Natale, Francesco.
Not Available
Not Available
The appeal of multiscale modeling approaches is predicated on the promise of combinatorial synergy. However, this promise can only be realized when distinct scales are combined with reciprocal consistency. Here, we consider multiscale molecular dynamics (MD) simulations that combine the accuracy and macromolecular flexibility accessible to fixed-charge all-atom (AA) representations with the sampling speed accessible to reductive, coarse-grained (CG) representations. AA-to-CG conversions are relatively straightforward because deterministic routines with unique outcomes are achievable. Conversely, CG-to-AA conversions have many solutions due to a surge in the number of degrees of freedom. While automated tools for biomolecular CG-to-AA transformation exist, we find that one popular option, called Backward, is prone to stochastic failure and the AA models that it does generate frequently have compromised protein structure and incorrect stereochemistry. Although these shortcomings can likely be circumvented by human intervention in isolated instances, automated multiscale coupling requires reliable and robust scale conversion. In this work, we detail an extension to Multiscale Machine-learned Modeling Infrastructure (MuMMI), including an improved CG-to-AA conversion tool called sinceCG. This tool is reliable (~98% weakly correlated repeat success rate), automatable (no unrecoverable hangs), and yields AA models that generally preserve protein secondary structure and maintain correct stereochemistry. We describe how the MuMMI framework identifies CG system configurations of interest, converts them to AA representations, and simulates them at the AA scale while on-the-fly analyses provide feedback to update CG parameters. Application to systems containing the peripheral membrane protein RAS and proximal components of RAF kinase on complex eight-component lipid bilayers with ~1.5 million atoms is discussed in the context of MuMMI.
With the growing complexity of computational and experimental facilities, many scientific researchers are turning to machine learning (ML) techniques to analyze large scale ensemble data. With complexities such as multi-component workflows, heterogeneous machine architectures, parallel file systems, and batch scheduling, care must be taken to facilitate this analysis in a high performance computing (HPC) environment. Here, we present Merlin, a workflow framework to enable large ML-friendly ensembles of scientific HPC simulations. By augmenting traditional HPC with distributed compute technologies, Merlin aims to lower the barrier for scientific subject matter experts to incorporate ML into their analysis. As a producer–consumer workflow model, Merlin enables multi-machine, cross-batch job, dynamically allocated yet persistent workflows capable of utilizing surge-compute resources. Key features of Merlin are a flexible HPC-centric interface, low per-task overhead, multi-tiered fault recovery, and a hierarchical sampling algorithm that allows for $\mathscr{O}$(N) task execution and $\mathscr{O}$(N ln N) task queuing to ensembles of millions of tasks. In addition to Merlin’s design, we test the algorithm’s performance in an HPC center and demonstrate the ability to enqueue 40 million simulations in 100 s, with a 30 millisecond per-task overhead that is independent of ensemble size. Finally, we describe some example applications that Merlin has enabled on leadership-class HPC resources, such as the ML-augmented optimization of nuclear fusion experiments and the calibration of infectious disease models to study the progression of and possible mitigation strategies for COVID-19.
Significance Here we present an unprecedented multiscale simulation platform that enables modeling, hypothesis generation, and discovery across biologically relevant length and time scales to predict mechanisms that can be tested experimentally. We demonstrate that our predictive simulation-experimental validation loop generates accurate insights into RAS-membrane biology. Evaluating over 100,000 correlated simulations, we show that RAS–lipid interactions are dynamic and evolving, resulting in: 1) a reordering and selection of lipid domains in realistic eight-lipid bilayers, 2) clustering of RAS into multimers correlating with specific lipid fingerprints, 3) changes in the orientation of the RAS G-domain impacting its ability to interact with effectors, and 4) demonstration that RAS–RAS G-domain interfaces are nonspecific in these putative signaling domains.
The advancement of machine learning techniques and the heterogeneous architectures of most current supercomputers are propelling the demand for large multiscale simulations that can automatically and autonomously couple diverse components and map them to relevant resources to solve complex problems at multiple scales. Nevertheless, despite the recent progress in workflow technologies, current capabilities are limited to coupling two scales. In the first-ever demonstration of using three scales of resolution, we present a scalable and generalizable framework that couples pairs of models using machine learning and in situ feedback. We expand upon the massively parallel Multiscale Machine-Learned Modeling Infrastructure (MuMMI), a recent, award-winning workflow, and generalize the framework beyond its original design. We discuss the challenges and learnings in executing a massive multiscale simulation campaign that utilized over 600,000 node hours on Summit and achieved more than 98% GPU occupancy for more than 83% of the time. We present innovations to enable several orders of magnitude scaling, including simultaneously coordinating 24,000 jobs, and managing several TBs of new data per day and over a billion files in total. Finally, we describe the generalizability of our framework and, with an upcoming open-source release, discuss how the presented framework may be used for new applications.
MuMMI RAS is the application component of the Multiscale Machine-Learned Modeling Infrastructure (MuMMI). It simulates RAS protein interactions at three scales of resolution coupled with ML-based selection and in-situ feedback. MuMMI RAS is particularly configured for massive scale, running thousands of simultaneous jobs with several terabytes of data on Lassen and Summit.
MuMMI Core is the central component of the Multiscale Machine-Learned Modeling Infrastructure (MuMMI), a scalable and generalizable framework that couples pairs of models using machine learning and in situ feedback across three scales of resolution.
This Python package is developed for converting GROMACS files to ddcMD inputs. The code makes it easier for ddcMD users to convert GROMACS MD simulations into ddcMD. It is used in the MuMMI framework for multiscale MD simulations.
Sina is a tool set for modern scientific data management that provides flexible, light-weight support of non-bulk data capture for retention in and queries against SQL and noSQL data stores. HPC simulations traditionally maintain their data in files. Extracting data of interest for subsequent analysis then requires the time-consuming process of traversing directories and scraping data from files in a variety of formats. Sina facilitates capturing relevant data during execution or post-processing of simulation runs for retention in and queries from a modern data store. The tools are sufficiently general to allow for the inclusion of new fields as scientists learn more about their data. Libraries, currently in C++ and Python, and a command line interface (CLI) are provided. Sina's flexibility starts with a general schema, in JSON, for the collection of non-bulk simulation data. JSON provides a flexible, human-readable representation of the data that of interest. Sina currently has a C++ library for simulations to write data to and read from a schema-compliant file for subsequent ingestion into one of the supported data stores. However, applications are free to write their data directly into a schema-compliant file. Python packages provide data ingestion, management, query, and export capabilities. A command line interface (CLI) provides simplified access to these features. A common application programming interface (API) is used to maintain and query data in any of the supported data stores, which are currently limited to SQL and Apache Cassandra (a column store). Tutorials, demonstrations, and examples illustrate aspects of the process using scripts and Jupyter notebooks.
Whether lipid rafts are present in the membranes of living cells remains hotly disputed despite their incontrovertible existence in liposomes at 298 K. In attempts to resolve this debate, molecular dynamics (MD) simulations have been extensively used to study lipid phase separation at high resolution. However, computation has been of limited utility in this respect because the experimental distributions of phases in lamellar lipid mixtures are poorly reproduced by simulations. In particular, all-atom (AA) approaches suffer from restrictions on accessible time scales and system sizes whereas the more efficient coarse-grained (CG) force fields remain insufficiently accurate to achieve correspondence with experiment. We refine the CG Martini parameters for the high- and low-melting temperature (T m ) lipids 1,2-dipalmitoyl-sn-glycero-3-phosphatidylcholine (DPPC) and 1,2-dioleoyl-sn-glycero-3-phosphatidylcholine (DOPC). Our approach involves the modification of bonded Martini parameters based on fitting to atomistic simulations conducted with the CHARMM36 lipid force field. The resulting CG parameters reproduce experimental structural and thermodynamic properties of homogeneous lipid membranes while concurrently improving simulation fidelity to experimental phase diagrams of DPPC, DOPC, and cholesterol lipid mixtures. Importantly, the refined parameters provide much better phase accuracy for regions near the critical point that mimic the lipid concentrations under physiological conditions.
MaestroWF is a Python tool and software package for loading YAML study specifications that represents a simulation campaign. The package is capable of parameterizing a study, pulling dependencies automatically, formatting output directories, and managing the flow and execution of the campaign. MaestroWF also provides a set of abstracted objects that can also be used to develop user specific scripts for launching simulation campaigns.