Coupling streaming AI and HPC ensembles to achieve 100–1000× faster biomolecular simulations

  title={Coupling streaming AI and HPC ensembles to achieve 100–1000× faster biomolecular simulations},
  author={Alexander Brace and Igor Yakushin and Heng Ma and Anda Trifan and Todd S. Munson and Ian T. Foster and Arvind Ramanathan and Hyungro Lee and Matteo Turilli and Shantenu Jha},
  journal={2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS)},
  • Alexander BraceI. Yakushin S. Jha
  • Published 10 April 2021
  • Computer Science
  • 2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS)
Machine learning (ML)-based steering can improve the performance of ensemble-based simulations by allowing for online selection of more scientifically meaningful computations. We present DeepDriveMD, a framework for ML-driven steering of scientific simulations that we have used to achieve orders-of-magnitude improvements in molecular dynamics (MD) performance via effective coupling of ML and HPC on large parallel computers. We discuss the design of DeepDriveMD and characterize its performance… 

Figures and Tables from this paper

Asynchronous Execution of Heterogeneous Tasks in AI-coupled HPC Workflows

An analysis of an important class of heterogeneous work, viz., AI-driven HPC workflows, to investigate asynchronous task execution requirements and properties and proposes key metrics that can be used to determine qualitative benefits when employing asynchronous execution.

Simulation-Based Parallel Training

The ongoing work to design a training framework that alleviates bottlenecks in numerical simulations by generating data in parallel with the training process and presenting a strategy to mitigate this bias with a memory buffer is presented.

The Ghost of Performance Reproducibility Past

This paper traces the experience uncovering performanceuctuations of ensemble applications, and unsuccessful attempts, so far, at trying to discern the underlying cause(s) of performance fluctuations, and ruminate over the implications for how to measure application performance.



Deep clustering of protein folding simulations

It is shown that the CVAE model can quantitatively describe complex biophysical processes such as protein folding, and can be used to learn latent features of protein folding that can be applied to other independent trajectories, making it particularly attractive for identifying intrinsic features that correspond to conformational substates that share similar structural features.

How Fast-Folding Proteins Fold

Results of atomic-level molecular dynamics simulations of 12 proteins reveal a set of common principles underlying the folding of 12 structurally diverse proteins that spontaneously and repeatedly fold to their experimentally determined native structures.

AI-Driven Multiscale Simulations Illuminate Mechanisms of SARS-CoV-2 Spike Dynamics

A generalizable AI-driven workflow is developed that leverages heterogeneous HPC resources to explore the time-dependent dynamics of molecular systems and demonstrates how AI can accelerate conformational sampling across different systems and pave the way for the future application of such methods to additional studies in SARS-CoV-2 and other molecular systems.

A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise

DBSCAN, a new clustering algorithm relying on a density-based notion of clusters which is designed to discover clusters of arbitrary shape, is presented which requires only one input parameter and supports the user in determining an appropriate value for it.

RADICAL-Cybertools: Middleware Building Blocks for Scalable Science

This paper provides an overview of RCT systems, their impact, and the architectural principles and software engineering underlying RCT.

#COVIDisAirborne: AI-Enabled Multiscale Computational Microscopy of Delta SARS-CoV-2 in a Respiratory Aerosol

This work dramatically extends the capabilities of multiscale computational microscopy to address the significant gaps that exist in current experimental methods, which are limited in their ability to interrogate aerosols at the atomic/molecular level and thus ob-scure the authors' understanding of airborne transmission.

Intelligent Resolution: Integrating Cryo-EM with AI-driven Multi-resolution Simulations to Observe the SARS-CoV-2 Replication-Transcription Machinery in Action

This study provides unprecedented access to study the SARS-CoV-2 RTC machinery, while providing general capability for AI-enabled multi-resolution simulations at scale, and develops an innovative workflow that bridges the gap between these resolutions.

Colmena: Scalable Machine-Learning-Based Steering of Ensemble Simulations for High Performance Computing

The design of Colmena is described and its capabilities are illustrated by applying it to electrolyte design, where it both scales to 65536 CPUs and accelerates the discovery rate for high-performance molecules by a factor of 100 over unguided searches.

Stream-AI-MD: streaming AI-driven adaptive molecular simulations for heterogeneous computing platforms

It is shown that Stream-AI-MD simulations can improve time-to-solution by ~50X for BBA protein folding, and performance trade-offs involved in implementing AI-coupled HPC workflows on heterogeneous computing architectures are discussed.