AI summary by TwinMind
Dmitry Korkin: Computational Biology of Coronavirus | Lex Fridman Podcast #90
Lex Fridman
2h 9m 2s
Introduction and Background
- Dimitri Korkin: Professor of bioinformatics/computational biology at WPI (Worcester Polytechnic Institute)
- Specializes in bioinformatics of complex diseases, computational genomics, systems biology, biomedical data analytics
- February work: Used viral genome of COVID-19 to reconstruct 3D structure of major viral proteins and their interaction with human proteins
- Created structural genomics map of coronavirus, made data open/available to researchers worldwide
Virus Intelligence and Evolution
- Views viruses as "villains that do their work so perfectly well"
- Viruses as intelligent machines: limited functions, minimal information, ability to self-modify
- Intelligence lies in simplicity: "ability to do so much with so little material and information"
- Swarm intelligence concept: viruses acting as cohesive whole
- Natural vs engineered pandemics: More concerned with naturally occurring viruses
- Evolutionary process creates efficiency through natural selection
Virus Contagiousness and Transmission
- R₀ (basic reproduction number) examples:
- COVID-19: 1.5-3.0
- Smallpox: 5-7
- Measles: ~15
- Factors affecting contagiousness:
- Transmission methods (aerosol, droplet size, airborne capability)
- Surface survival ability
- Replication speed
- Incubation period
- Asymptomatic transmission percentage
- Iceland study: 50% asymptomatic cases
- Butterfly effect: Small genomic changes enable species jumping (bat to human to human-to-human)
COVID-19 Infection Mechanism
- Virus attachment via spike protein (trimer formation) to ACE2 human receptor
- COVID-19 attaches more efficiently to ACE2 than SARS
- Steps: attachment → membrane fusion → RNA release → hijacking cellular machinery
- Virus lacks ribosomes, must hijack host ribosomes for protein synthesis
- Four main structural proteins:
- Spike protein (attachment)
- Membrane protein (envelope formation, highest copy number)
- Envelope protein (function still being researched)
- Nucleocapsid protein (RNA protection)
- 29 total proteins vs 8-9 in influenza
Protein Function Discovery Process
- Computational approach: Compare new proteins to known protein databases
- Homology/template-based modeling
- Similar sequences → similar structures → similar functions
- Generate predictions for experimental validation
- Challenges with truly novel proteins requiring ab initio methods
Protein Folding
- Million-dollar challenge problem
- Proteins fold efficiently into canonical conformations
- Most proteins have 2-3 structural domains
- Current success: proteins up to 100 residues
- Difficulties: transmembrane proteins, larger multi-domain proteins
- Machine learning approaches emerging (AlphaFold by DeepMind)
- CASP competition: Olympic games for protein folders
- Folding@Home: Crowdsourced protein folding, players achieved better results than scientists
Protein Data Bank and Open Science
- Community effort: scientists submit solved structures
- No hoarding culture in structural biology
- Pandemic accelerated collaborative efficiency
- Preprint culture changing knowledge sharing
- Knowledge becoming core value over journal publication
Research Paper: SARS-CoV-2 Structural Analysis
- Origin story: September 2019 Triple E virus outbreak in New England
- Eastern equine encephalitis virus
- Brain-targeting, mosquito-transmitted
- 30%+ mortality rate
- Local schools closed, outdoor activity restrictions
- Class project studying Triple E using bioinformatics
- December 2019: Projects completed
- January 2020: Applied same methodology to novel coronavirus
Methodology
- Started with genome sequence (30,000 nucleotides, 29 genes)
- Used SARS protein information for gene boundary identification
- 3D structure prediction via homology modeling
- Identified protein complexes and interactions
- Created interactome predictions
Key Findings
- Pattern: Some proteins identical to SARS/bat coronaviruses, others heavily mutated
- Evolution not uniform across viral genome
- Drug binding sites largely intact → existing antivirals potentially effective
- Structural differences clustered in 3D space → functional significance
- Viral particle modeling: ~100nm diameter, ~5,200 spike proteins, 200-400 membrane protein dimers
Antiviral Drug Development
- Two approaches: vaccines (prevention) vs antivirals (treatment)
- Remdesivir most promising: RNA polymerase inhibitor, originally for Ebola/Marburg
- Other candidates: protease inhibitors, repurposed malaria drugs
- Computational predictions guide experimental drug screening
Universal Vaccines
- Current seasonal flu shots: arms race with constantly changing virus
- Goal: vaccine effective against all strains regardless of mutations
- Would prevent future pandemic strains
Agent-Based Epidemiological Modeling
- 5-year project: "Zombies on a Cruise Ship"
- Confined environment modeling (cruise ships, schools, companies)
- Novel approach: explicit pathogen agents alongside human agents
- Parameters: transmission methods, surface survival, viral shedding rates, asymptomatic periods
- Diamond Princess provided real-world validation opportunity
- Modeled various viruses: Norovirus, Ebola, flu, SARS, fictional MEV-1 from Contagion movie
COVID-19 Unique Characteristics
- Long asymptomatic period creates new containment challenges
- Asymptomatic individuals shed sufficient virus for transmission
- First week most contagious period
- Tears not contagious (recent finding)
- Mask efficacy: protects others more than wearer
Treatment Outlook
- Antivirals likely short-term solution over social distancing
- Mutation concerns: virus may develop drug resistance
- 18-month vaccine timeline historically very accelerated (normally ~10 years)
- Regulatory acceleration vs safety balance
Personal Background
- Born in Russia, moved to Canada then US in 1999 at age 22
- Moscow State University mathematics background
- Childhood dream: passenger jet pilot → cybernetics → applied mathematics
- Strong Russian bioinformatics collaborations continue
- Science bobblehead collection: Newton, Einstein, Curie, Tesla, Darwin, Franklin
- Missing desired additions: Kolmogorov, Turing, von Neumann, Linus Pauling
Philosophical Reflections
- Virus research reveals human fragility
- Single tiny entity can impact entire human civilization
- Reinforces need for societal cooperation
- Provides hope that scientific work has meaningful impact
- Wilson quote: Viral genetic diversity likely exceeds all other life combined
