Preskoči na sadržaj

The challenges of the upcoming exascale supercomputing era in computational biochemistry

Dr. Vedran Miletić (

😎 Group for Applications and Services on Exascale Research Infrastructure, Faculty of Informatics and Digital Technologies, University of Rijeka

Research Class, FIDIT, UniRi, 26th January 2022

Stream and recording check

  • OBS
  • BBB

Dr. Vedran Miletić's previous research work

  • Dr. Branko Mikac's group at FER Dept. of Telecommunications
  • What to do after finishing the Ph.D. thesis? 🤔
    • NVIDIA CUDA Teaching Center (later: GPU Education Center)
    • research in Dr. Željko Svedružić’s Biomolecular Structure and Function Group and Group (BioSFGroup,
  • postdoc in Dr. Frauke Gräter's Molecular Biomechanics (MBM) group at Heidelberg Institute for Theoretical Studies
    • collaboration with GROMACS developers from KTH, Max Planck Institute for Biophysical Chemistry (now: Multidisciplinary Sciences), and University of Virginia

RxTx Research

  • returned from Heidelberg, became a Senior Lecturer
    • 90% working hours teaching (courses + Bura supercomputer), 10% administration, 0% research
  • started RxTx Research (
    • collaboration with Patrik Nikolić (, former student researcher in BioSFGroup)
    • vision: advancing the pharmaceutical drug research by improving the scientific software behind the scenes
    • developed open-source high-throughput virtual screening engine RxDock (, until promotion to assist. prof.)

Group for Applications and Services on Exascale Research Infrastructure (GASERI)

  • The main interest: the application of exascale computing to solve problems in computational biochemistry
  • The goal: design better-performing algorithms and offer their implementations for academic and industrial use to
    • study the existing molecular systems faster
    • study the existing molecular systems in more detail
    • study larger molecular systems


  • a supercomputer is a computer with a high level of performance as compared to a general-purpose computer
    • also called high performance computer (HPC)
  • measure: floating-point operations per second (FLOPS)
    • PC -> teraFLOPS; Bura -> 100 teraFLOPS
    • modern HPC -> 1 do 10 petaFLOPS, top 442 petaFLOPS
    • future exascalar HPC -> 1+ exaFLOPS
  • nearly exponential growth of FLOPS over time (source: Wikimedia Commons File:Supercomputers-history.svg)

bg 80% Computing power of the top 1 supercomputer each year, measured in FLOPS

More heterogeneous architectures require complex programming models

  • different types of accelerators
    • GPUs (half, single, double precision), TPUs/TCGPUs, FPGAs
    • in-network and in-storage computation (e.g. BlueField DPU)
  • several projects to adjust existing software for the exascale era
    • Software for Exascale Computing (SPPEXA)
    • Exascale Computing Project (ECP)
    • European High-Performance Computing Joint Undertaking (EuropHPC JU)


  • full title: Unified Long-range Electrostatics and Dynamic Protonation for Realistic Biomolecular Simulations on the Exascale
  • principal investigators:
    • Helmut Grubmüller (Max Planck Institute for Biophysical Chemistry, now Multidisciplinary Sciences)
    • Holger Dachsel (Jülich Supercomputing Centre)
    • Berk Hess (Stockholm University)
  • molecular dynamics visualization: Electron transport chain


The particle mesh Ewald method (PME, currently state of the art in molecular simulation) does not scale to large core counts as it suffers from a communication bottleneck, and does not treat titratable sites efficiently.

The fast multipole method (FMM) will enable an efficient calculation of long-range interactions on massively parallel exascale computers, including alternative charge distributions representing various forms of titratable sites.

SPPEXA Projects - Phase 2 (2016 - 2018)

Planned GROMACS developments (1/2)

  • heterogeneous parallelism presently uses GPUs, could be expanded to also use DPUs
    • custom-silicon Anton 2 supercomputer's hardware and software architecture could be an inspiration
    • identification of packets that do not need to be delivered to all receivers and force reductions
    • NVIDIA already offers free developer kits to interested parties for similar purposes

Planned GROMACS developments (2/2)

  • molecular dynamics simulations are periodic
  • simulation box types: cubic, rhombic dodecahedron
  • present design and implementation of the fast multipole method only supports cubic boxes
    • it is possible to also support rhombic dodecahedron: ~30% less volume => ~30% less computation time per step required
  • potentially apply for HrZZ UIP (if announced)

Potential GROMACS developments

  • Monte Carlo (Davide Mercadante, University of Auckland)
    • many efforts over the years, none with broad acceptance
    • should be rethought, and then designed and implemented from scratch with exascale in mind
  • polarizable simulations using the classical Drude oscillator model (Justin Lemkul, Virginia Tech)
    • should be parallelized for multi-node execution
  • other drug design tools such as Random Acceleration Molecular Dynamics (Rebecca Wade, Heidelberg Institute for Theoretical Studies and Daria Kokh, Cancer Registry of Baden-Württemberg)

Interesting developments in the broader computational biochemistry ecosystem

RDKit and RxDock



  • protein structure != protein sequence
  • earlier computational solutions: Folding@home
  • enabled by the evolution of GPUs and developments in AI
  • Forbes calls it The Most Important Achievement In AI—Ever: 'Critical Assessment of Protein Structure Prediction co-founder and long-time protein folding expert John Moult put the AlphaFold achievement in historical context: "This is the first time a serious scientific problem has been solved by AI."'

Potential development: HTVSDB

  • web interface and REST API to a molecular database and molecular docking service
  • open-source software so it could be hosted locally by other research groups at other universities
  • unique features: molecular recommendation, federation
  • based on RDKit, RxDock, and potentially AlphaFold
  • long-term evolution on a best-effort basis

bg Drug Discovery and Development Pipeline

Figure source: Cui W, Aouidate A, Wang S, Yu Q, Li Y and Yuan S (2020) Discovering Anti-Cancer Drugs via Computational Methods. Front. Pharmacol. 11:733. doi: 10.3389/fphar.2020.00733

Unified vision and specific applications

  • high-throughput virtual screening and molecular dynamics simulations could be offered as a service to Croatian, regional, and EU research groups
    • methods -> algorithms -> applications
  • e.g. industry/academic group has a molecular target
    • RxDock, RDKit (HTVSDB, KNIME/Python automation): millions of molecules -> tens of molecules
    • GROMACS (KNIME/Python automation) -> tens of molecules -> several molecules

Author: Vedran Miletić