Toxicology used to be about animal counts and LD₅₀ trivia. In 2025 it’s a software‑first, systems‑biology discipline colliding with the real world of regulators and drug developers. The Society of Toxicology’s Annual Meeting in Orlando showed a field that’s both more computational and more translational—closer to the clinic, closer to manufacturing, and closer to regulators’ inboxes.
Four shifts you could feel in the sessions
- NAMs are getting operational. Cell‑based assays, high‑content imaging and adverse outcome pathway (AOP) frameworks are now wired into screening cascades. The question is no longer “if” but “which NAM for which decision?”
- Microphysiological systems (MPS)—organ‑on‑chip, multi‑organ platforms—are earning their keep in DILI and cardiotox risk, especially when paired with omics readouts and longitudinal imaging. The trick is standardization: vendors who showed robust assay performance metrics got the crowds.
- PBPK meets QIVIVE in the mainstream. Exposure modeling is moving earlier in discovery and later into clinical supporting evidence, tying together in vitro data with human relevance.
- AI and knowledge graphs are cleaning up the literature mess—normalizing ontologies, flagging inconsistent endpoints, and surfacing read‑across candidates. The grown‑up teams arrived with model cards, drift monitoring and transparent training sets.
Hot topics that spilled into the hallways
- Nitrosamines and extractables/leachables: manufacturing controls, sensitive analytics, and risk evaluation playbooks dominated standing‑room sessions.
- PFAS and microplastics: regulators are leaning into precaution; industry wants exposure‑based thresholds and better fate/transport models.
- Occupational exposure in battery and semiconductor supply chains: expect new guidance on chronic low‑dose risks.
What regulators want (and will reward)
- Decision‑grade validation: inter‑lab reproducibility and performance envelopes, not one‑off heroic datasets.
- Transparent uncertainty: show the sensitivity analysis and where your model breaks.
- Lifecycle thinking: how will your method support post‑marketing safety and change control, not just INDs?
Startup and vendor playbook
- Sell workflow, not widgets: ELNs, LIMS, data standards and audit trails that make NAMs usable at scale.
- Be OECD‑test‑guideline‑literate and map your assays to AOPs and regulatory decisions.
- Offer SaaS + services packages that help teams convert old data into machine‑readable, ontology‑aligned assets.
Why this matters beyond the tox community
As biology goes software‑defined, safety becomes a data problem. The companies that win will treat toxicology as a continuous signal across R&D and manufacturing, not a gate at the end. That means better drugs faster—and safer consumer products by design.
Dates & place: 16–20 March 2025, Orlando, Florida.
