#90 – Dmitry Korkin: Computational Biology of Coronavirus

Episode Summary

Executive Summary: Dimitri Corkin explains how viruses work as minimal, highly efficient biological machines, and why bioinformatics is essential for understanding and countering them. The conversation centers on SARS-CoV-2 structural genomics, protein folding, virus-host interactions, drug and vaccine design, and pandemic epidemiology, emphasizing that open data, computational modeling, and rapid collaboration can accelerate discovery.

Main Topics: Viruses as efficient biological machines (Priority: 5/5): Corkin frames viruses as simple but remarkably optimized systems that do a lot with very little genetic material, stressing both their elegance and their threat. COVID-19 structural genomics and protein modeling (Priority: 5/5): He describes his group’s work mapping SARS-CoV-2 proteins, predicting structures, and identifying conserved versus mutated regions to support drug discovery. Virus infectivity, mutation, and pandemic risk (Priority: 4/5): The discussion covers how viruses jump species, why contagion varies, and why naturally occurring pandemics remain the biggest concern. Protein folding and computational biology (Priority: 4/5): Corkin explains folding as a hard computational problem, discusses template-based and ML approaches, and notes progress from initiatives like AlphaFold and CASP. Vaccines, antivirals, and therapeutic strategy (Priority: 4/5): He distinguishes prevention via vaccines from treatment via antivirals, citing mechanisms such as protease inhibition and polymerase disruption. Agent-based epidemiological simulation (Priority: 3/5): He discusses modeling outbreaks on cruise ships and other closed environments using agents for both hosts and pathogens to test interventions. Open science and collaborative research culture (Priority: 3/5): The episode highlights preprints, shared databases, and rapid collaboration as key accelerants during the pandemic.

Key Arguments: Viruses are best understood as highly efficient machines that exploit host machinery; their power comes from simplicity, not complexity. Naturally occurring pandemics are the main long-term concern, because evolution continuously generates new strains and cross-species jumps. Bioinformatics can rapidly infer viral protein structure and function by comparing new sequences to known proteins, then guide experimental validation. Protein folding remains a major unsolved computational problem, though machine learning and community benchmarks are improving performance. Structural conservation across related viruses can identify druggable sites that may still work in a novel virus, as seen with SARS-CoV-2. Vaccines and antivirals require different strategies: vaccines train immunity before infection, while antivirals block critical viral functions after infection. Agent-based models are useful because they can incorporate host behavior, pathogen properties, asymptomatic shedding, and surface survival to compare interventions. Open data and preprints have become central to fast-moving scientific response, shifting value from journals to the underlying knowledge itself.

Data Points: Coronavirus receptor binding: ACE2 - Corkin says SARS-CoV-2 attaches to the human ACE2 receptor, similar to SARS. Estimated COVID-19 R0: 1.5 to 3 - He cites estimates for how many others one infected person may spread the virus to. Smallpox R0: 5 to 7 - Used as a comparison to show why smallpox was much more contagious than COVID-19. Measles R0: 15 and up - Referenced as an example of an even more contagious virus. H7N9 mortality rate: above 30% - Mentioned as a highly lethal avian flu strain that was not pandemic. Coronavirus proteins: at least 29 - He contrasts SARS-CoV-2’s protein count with influenza’s smaller protein set. Influenza proteins: around 8 or 9 - Used for comparison with the larger coronavirus genome. Virion size: around 80 nanometers - Approximate size of a single SARS-CoV-2 particle. Spike proteins per virion: 50 to 100 - Estimated number of trimeric spike proteins on an average particle. Membrane protein dimers per virion: 200 to 400 - Approximate number arranged in a lattice on the viral surface. Asymptomatic cases: about 30% or more - He notes a substantial fraction of infected people may show no symptoms while still shedding virus. Genome length: roughly less than 30,000 nucleotides - Describes the SARS-CoV-2 RNA genome used to identify genes and proteins. Protein folding benchmark: small proteins up to 100 residues - He says computational methods are strongest on relatively small proteins. Vaccine development timeline: up to 10 years historically - Used to illustrate how accelerated COVID-era timelines are. Paper timing: 5 or 6 days - The preprint took several days to be screened and posted during the early pandemic.

Pivotal Quotes: "The virus itself, I mean, it's not the limiting organism. It's a machine to me." — Dimitri Corkin: He explains his conceptual model of viruses as optimized biological machines. "I mean, this is perhaps my favorite example of a butterfly effect because it's really, I mean, it's even tinier than a butterfly." — Dimitri Corkin: He describes how small mutations can enable cross-species transmission and trigger massive outbreaks. "I think now the knowledge is becoming sort of the core value, not the paper or the journal where this knowledge is published." — Dimitri Corkin: He reflects on how preprints and rapid sharing are changing scientific culture during the pandemic.

Implications: The episode shows how computational biology can shorten the path from genome to therapeutic insight. For listeners, it underscores the value of open data, rapid collaboration, and vigilance about mutation-driven future pandemics.