Skip to content

Sergei L. Kosakovsky Pond — "Interactive Genomic Analytics with responsive notebooks, in-the-browser computation, and AI-assisted development"

Affiliation: Temple University — Institute for Genomics and Evolutionary Medicine Day / session: Day 2 (Tue Jun 24, 2025) — Keynote 1, Grace Auditorium (9:00–10:00) Talk title: "Interactive Genomic Analytics with responsive notebooks, in-the-browser computation, and AI-assisted development" Slides (Zenodo DOI): to verify after publish Video: GBCC2025 playlist Abstract / schedule: GBCC2025 program

Thesis

Genomic analysis is migrating off the server and into the browser — responsive notebooks running compiled code via WASM (Pyodide / WebR / native HyPhy builds), with LLM agents drafting the analysis scaffolding around them. The argument is almost certainly that the combination of in-browser compute, reactive notebook UI, and AI-assisted authoring collapses the distance between a published method and a non-developer using it on their own data — i.e., what Datamonkey did for selection analysis, generalized.

Speaker context

Pond is the lead developer of HyPhy (likelihood-based phylogenetic inference for selection, recombination, and rate variation) and a long-time contributor to Datamonkey, the public web front-end that has made HyPhy's selection-analysis methods accessible to non-developers for two decades. The Day-2 opening keynote slot is well chosen: Datamonkey is one of the older "compute-on-our-server" web tools in genomics, so a talk from its lead author about moving compute into the user's browser is a credible signal that the architecture is shifting. Beyond the title, the specific tooling — Pyodide vs. WebR vs. a HyPhy-native WASM build — is not yet knowable from the schedule alone.

Connections to the corpus

  • AACR axis: agentic-ai — "AI-assisted development" puts this talk in conversation with the LLM-agent thread running through Day 2's Oral Session 1
  • GAIA — Galaxy's AI agent framework; same Day-2 morning, sets the AI tone for the whole conference
  • GalaxyMCP — the MCP-server-for-Galaxy entry; complementary "AI meets workflow tooling" angle
  • Cross-link to the GBCC2025 index for the full Day-2 oral lineup that this keynote opens

Open questions

  • Which WASM runtime does Pond actually demo — a HyPhy-native build, Pyodide, WebR, or all three? The browser-compute story changes meaningfully depending on whether "in-the-browser" means a port of HyPhy or a generic Python/R kernel.
  • How concrete is the AI-assisted development claim — a working notebook agent, or a forward-looking sketch? Datamonkey-as-an-LLM-tool would be a natural next step but isn't promised by the title.
  • Does the talk address the deployment / privacy argument for browser-side compute (no server, no upload of patient data) explicitly, or stay on the developer-experience side?