AACR Annual Meeting 2026

This is not a racetrack. It is the tool layer: spatial omics to see how cells sit together, molecular glues to trick “undruggable” proteins into degradation, AI to turn a slide from a glance into a measurement. Low hallway heat is not a demotion — the things nobody is posting are often the next five years of infrastructure.

Three names from the floor: Amphista’s AMX-883 (a BRD9 glue) shown for the first time; 10x’s in situ spatial platform; Stanford’s Biomni wiring literature, experiments and code into an agent. They will not out-heat daraxonrasib. The next layer of KRAS / ADC / MRD still runs on them.

Glues and PROTACs both trick proteins into degradation; glues are smaller and more oral, and discovery still looks like luck. New Drugs on the Horizon put AMX-883 in a first-disclosure seat. Spatial omics writes “who sits next to whom” as a quantifiable input, not only a pretty picture.

Pathology is moving from reading to measuring: TIL / TLS become continuous variables, then get tied to survival and IO response. Biomni, a 25.6-million-cell foundation model, and TLS detection at AUC 0.92 are three scales of the same sentence: AI is no longer a one-task demo.

• 1KRAS / RAS targeting: from platform to Phase 3
• 2PD-1 × VEGF bispecifics after ivonescimab
• 3Next-generation ADCs: payload innovation
• 4Immunotherapy: CAR-T in solid tumors, mRNA vaccines, TCR-T
• 6MRD / ctDNA / DTP: from detection to treatment decisions
• 8Medicinal chemistry beyond KRAS: CDK2, WEE1, SERD
• 10Breast cancer: HER2 de-escalation, MRD, TNBC
• 11Lung cancer: perioperative IO and KRAS NSCLC
• 12The rest: ovarian, HNSCC, GBM and other floor tumors