Interview
Dmitry Korkin: Computational Biology of Coronavirus | Lex Fridman Podcast #90
- Dimitri Korkin is a professor of bioinformatics and computational biology at Worcester Polytechnic Institute (WPI), specializing in the computational genomics and systems biology of complex diseases.
- Korkin's research group utilized the SARS-CoV-2 viral genome in February to reconstruct the 3D structures of major viral proteins and their interactions with human proteins, creating an open-source structural genomics map.
- Viral Nature: Korkin views viruses not as living organisms but as "intelligent machines" optimized for efficiency, possessing just enough information to execute necessary functions and modify themselves.
- Evolutionary Intelligence: The "intelligence" of a virus is attributed to its simplicity, maximizing functionality with minimal material, potentially representing a form of basic swarm intelligence.
- Pandemic Risks: Korkin is more concerned with naturally occurring pandemics (e.g., new influenza or coronavirus strains) than engineered ones, citing the historical precedent of natural emergence.
- Smallpox Context: Historically, smallpox was the most dangerous virus, with an estimated basic reproduction number ($R_0$) of 5–7, compared to the estimated $R_0$ of 1.5–3 for SARS-CoV-2.
- Transmissibility Factors: Viral contagion is determined by a combination of biological factors (aerosol transmission, surface survival, replication speed, incubation period) and social factors.
- Asymptomatic Spread: A study in Iceland suggests approximately 50% of SARS-CoV-2 cases are asymptomatic, yet these individuals remain contagious and shed the virus.
- Viral Mutation: Small changes in viral genomes (mutations of a few atoms) can enable cross-species jumping (e.g., bat to human) and human-to-human transmission.
- Viral Engineering Risks: Modifying viruses to increase transmissibility is theoretically possible (as demonstrated with influenza studies), but uncontrolled engineering could lead to mutual self-destruction due to lack of containment.
- Universal Vaccine Goal: The objective is to develop a vaccine effective against all strains of a virus (e.g., Influenza A) regardless of the species it jumped from, ending the "arms race" with constantly mutating strains.
- Pathogenicity vs. Contagion: There appears to be a biological trade-off where highly pathogenic viruses (e.g., MERS with >30% mortality) tend to be less contagious than less lethal ones, suggesting an evolutionary balance that does not necessarily aim to eradicate the host.
- SARS-CoV-2 Structure: The virus contains 29 proteins, significantly more than influenza (8–9), encoded by an RNA genome of roughly 30,000 nucleotides.
- Attachment Mechanism: SARS-CoV-2 attaches to human cells via the ACE2 receptor using a trimeric spike protein, with evidence suggesting it binds more efficiently than the original SARS virus.
- Protein Complexity: The viral envelope includes spike proteins (functional units), membrane proteins (maintain curvature), and envelope proteins (function currently unknown); internal ribonucleoproteins protect the RNA.
- Bioinformatics Function: Computational methods infer protein function by comparing sequences to known databases (homology modeling); for novel proteins, ab initio folding prediction is required but remains challenging.
- Protein Folding: Determining 3D protein structure is a complex combinatorial problem; while tools like "Foldit" (a game) and DeepMind's AlphaFold have achieved high accuracy, robust ab initio prediction for large proteins remains an ongoing challenge.
- Open Data Culture: The Protein Data Bank (PDB) has served as a global repository for decades, facilitating a culture where structural data is openly shared to accelerate prediction and discovery.
- Research Acceleration: The pandemic catalyzed an unprecedented level of scientific collaboration, with pre-prints and rapid review cycles replacing traditional, slower publication timelines.
- Drug Repurposing: Structural analysis revealed that binding sites for known small-molecule drugs (e.g., against SARS and MERS) remain largely intact in SARS-CoV-2, suggesting existing drugs like Remdesivir may be effective.
- Agent-Based Modeling: Korkin's team developed simulations originally for "zombies on a cruise ship" (Norwalk virus) that were rapidly adapted to model SARS-CoV-2, incorporating the virus itself as a distinct agent.
- Modeling Insights: Simulations highlight the critical impact of asymptomatic periods and viral shedding rates on containment strategies, noting that long asymptomatic infectious periods complicate traditional intervention methods.
- Mask Efficacy: Masks are effective in preventing asymptomatic carriers from spreading the virus to others, even if they offer less protection to the wearer.
- Korkin's Background: Born in Russia, Korkin holds a background in mathematics and cybernetics, having studied at Moscow State University before moving to the US in 1999.
- Scientific Community: While noting strong bioinformatics collaborations in Russia, Korkin observes a lack of major AI conferences and significant barriers to travel for scientists from that region.
- Science Collection: Korkin collects scientific bobbleheads, including Watson, Crick, Marie Curie, Tesla, and his favorite, Rosalind Franklin, citing her as a pivotal figure in structural biology.
- Philosophical Outlook: Studying viruses has highlighted the fragility of human life and the necessity of societal cohesion, while reinforcing the value of scientific research in protecting humanity.