Skip to content

Session 9 — AI prediction of TCR–antigen binding

Series: 2026 SITC Computational IO Series, session 9 of 10 Date: November 9, 2026, 1:00–2:30 p.m. ET (90 min) Moderator: Marshall Thompson, PhD — The Rare Cancer Research Foundation Faculty:

  • Tao Wang, PhD — University of Texas MD Anderson Cancer Center
  • Li Zhang, PhD — University of California, San Francisco
  • Phil Bradley, PhD — Fred Hutchinson Cancer Center

Recording: SITC OnDemand after delivery Status: Upcoming

What the session covers

Machine-learning prediction of which T-cell receptors bind which antigens — the central unsolved problem in repertoire immunology, and the one that gates neoantigen-directed therapy. Three faculty across 90 minutes suggests multiple modelling approaches presented side by side; Bradley's group has long worked on TCR structural and sequence modelling. Inferred from the roster and slot length; confirm after delivery.

Why it matters to this corpus

  • The closest thing in the series to a benchmark-and-baselines talk, on a problem where published performance claims have historically been fragile — exactly the critical-reading posture the corpus takes in Why linear baselines win and Reading an FM paper critically.
  • Protein/receptor sequence models — same technology family as protein-LM explainability.
  • Completes the antigen pipeline the series builds across session 3 (repertoire readouts), session 5 (immunogen design), and session 7 (neoantigen-directed therapy).
  • Timing note: this session lands November 9, the week after the SITC 41st Annual Meeting (Nov 4–8, Phoenix) — expect overlap with material presented there.

Open items

  • [ ] Capture the recording once posted.
  • [ ] Capture all three talk titles and the specific models/benchmarks each presented.
  • [ ] Record which datasets were used — benchmark provenance is the crux of this field's reproducibility problem.

Sources