Accelerating the Discovery of DNA Based Materials using High-Performance Computing and Structural Bi
Abstract
This abstract is publicly releasable. Program Manager: Laura Kienker [342]. Structured assemblies of DNA and RNA molecules offer the, unique property to engineer biomaterials with nanoscale precision. Such nanomaterials have great potential for innovations in molec,ular data storage and quantum computing, excitonics, photonics, and therapeutics. The Bathe lab has pioneered algorithms for the rap,id sequence design of arbitrarily shaped 2D (PERDIX and METIS) and 3D (DAEDALUS and TALOS) DNA nanostructures. We have also implemen,ted scalable methods for the synthesis of designer DNA scaffolds and established the production of large-scale libraries of function,alized DNA nanostructures, including high-throughput assembly and characterization of DNA nanomaterials. Arranging functional molecu,les discretely on DNA and RNA enables the fabrication of nanomaterials with tailored physical, biochemical, and structural propertie,s. While the Bathe lab has recently developed a high-throughput assembly and characterization pipeline for functional DNA and RNA na,nomaterials, screening through the very large structural and chemical/biochemical space of possible structures is prohibitively time,-consuming and expensive. In silico screening tools can circumvent the experimental bottleneck by filtering through an extensive lib,rary of possible DNA and RNA nanomaterials design space using atomistic-level molecular simulations and state-of-the-art machine lea,rning approaches. Such calculations would require high performance computing (HPC) facilities with state-of-the-art graphics cards (,GPUs). Besides using computation to predict functional DNA and RNA nanomaterials, dedicated HPC facilities are essential for charact,erizing synthesized structures by cryogenic electron microscopy (cryoEM), which would minimally require at least a terabyte of stora,ge space for each nanostructure for structural characterization. Here we propose to build an HPC system that features 1.5 Petabytes,of storage, 16 Nvidia A100 GPUs, and 192-core AMD CPUs. The proposed HPC system will replace our current eight-year-old HPC cluster,with outdated hardware, particularly the GPUs, and complement our high-throughput experimental synthesis pipeline as well as modern,cryoEM facilities at MIT.nano. The computing cluster will accelerate research projects currently funded by the ONR, DOE, NIH, and NS,F. We will also provide access to the proposed HPC system to users of the cryoEM facility at MIT.nano, as well as the MIT Biological, Engineering Teaching Laboratory to provide students practical learning opportunities to use state-of-the-art computational methods,involving molecular simulation, machine learning, and structural analysis using cryoEM.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 01, 2022
- Source ID
- N000142212317
Entities
People
- Mark Bathe
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy