355 - Supercomputing for COVID-19

omega tau science & engineering podcast

Guests: Andrew Emerson, Daniel Jacobson
Host: Markus Voelter Shownoter: Andy Joiner

In this episode we look at how supercomputers are used to help with managing the pandemic. It’s a double-header with two guests. We start with Cineca‘s Andrew Emerson. As part of the EXSCALATE 4 COV EU-funded research project, he works of virtual screening of existing drugs regarding their potential efficacy against SARS-CoV-2. In part two we talk with Dan Jacobson of the Oak Ridge National Laboratory. He and his team used a big data analysis to understand how the virus “works”, and they figured out very interesting mechanisms and pathways.

Introduction

00:03:14

Andrew Emerson | Cineca Supercomputer Center | Exscalate4Cov Project | Computer Aided Drug Design | Simulation | Modelling | Synthesize | Side Effects | Active Ingredient | COVID-19 | SARS-CoV-2 | Virus | Drug Repurposing/Drug Repositioning | Toxicology | Virtual Screening | Protein | Enzyme | Cell | Replication | Inhibit | DNA | Cell Membranes | Amino Acid | RNA | Gene | Gene Expression | Genome | Lock and Key Model | Screening | Virtual Screening | Dompé | Parallel Process | Abstraction (PDB - Protein Data Bank) | Bonding | Experimental data (Proteins - accurate 3d data | Drug molecules - basic data) | X-Ray Diffraction | NMR | Cryogenic Electron Microscopy | Electromagnetic Force | Van der Waals Force | Electrostatic Interaction | LiGen | Molecular Docking | Toxicology | Few seconds per molecule for virtual screening

Molecular Dynamics Simulation

00:31:06

Molecular Dynamics Simulation | Thermal effects | Stable State | Statistical techniques | Numerical Simulation | Newton's Laws of Motion | Bond vibration | Coulomb's law | Multi-Body Problem | Unfold | Molecular dynamics simulation can take about a week | Clustering | Degrees of Freedom | Energy Surface | Boltzmann Distribution | Virtual screening filtered ~400,000 possible drugs to ~7000 to go to the laboratory | Vaccine | Immune Response | ~100 of ~7000 screened molecules were interesting | In Vitro Tests | 40 effective molecules limited virus replication | Antiviral | Cortisone

Computer Software and Hardware

01:09:44

GROMACS | Open Source | GPU | Homology Model | C++ | CUDA | Marconi 100 | Galileo Linux Cluster | Eni | HPC5 | Nvidia V100 | Scheduling

Results

01:20:24

Raloxifene | Clinical Trials | Dimer | Protease | Physical Chemistry | Computational Chemistry | Vector (Biology) | Vector (Mathematics)

Bradykinin Storm Analysis

01:30:59

Dan Jacobson | Computational Systems Biologist | Oak Ridge National Laboratory (OLCF - Oak Ridge Leadership Computing Facility) | Evolution | Pathogenicity | 3D Structure | Human pathogenesis | Demographic factors | Bradykinin Storm Paper | 17000 samples across 57 different tissues | Regulatory circuits | Bronchoalveolar Lavage fluid | Bronchoscope | Terminal Bronchioles | Luminal surface | Gas Exchange | Pathways (RAS - Renin-Angiotensin System) | Blood Pressure | Inflaitory RResponses | ACE2 | Cell Entry | ACE | Proteolytic Enzyme | Proteolytic cleavage | Peptide | Hypertensive | Upregulated (Ang-(1-9) - Angiotensin (1-9)) | AGTR2 receptor | Receptor | Bind | Kallikrein Kinin Pathway | C1-Inhibitor Protein | Biosynthesis | Vasodilation | Hyperpermeability | Plasma | Capillaries | Immune Cells | Transcription | Hyaluronic Acid | Hydrogel | Tropism | Renin | Vitamin D | Catabolise | Metabolite | Expression | Calcifediol | Icatibant | Dexamethasone | Phospholipase A2 | Arachidonic Acid | Cascade | Chemotaxis | Combinatorial Therapies

Computational Tools

01:56:58

Machine Learning | Functional Inference | Explainable AI | Data Analytics | ChIPseek | Transcription Factors | Traits | Phenotypes | Combinatorial Space | P-Values | Permutation Tests | Neural Network | Black Box | Classification Problem (IRF - Iterative Random Forests) | Decision Tree | Iterative Approach | Summit Supercomputer | Flash Memory | Gordon Bell Prize 2018 | Tensor Cores | Matrix Multiplication | Exascale Barrier | Hyperparameter Sweeps | Monte Carlo Simulation | R | Python | Parsers