Insilica CEO Luechtefeld将在AAPS 2026 PharmSci 360谈可审计智能体AI拓展计算毒理学

At PharmSci 360, Insilica CEO Thomas Luechtefeld will present on auditable agentic AI systems and how they are expanding computational toxicology beyond single predictive models.

In an interview with BioPharm International® ahead of the American Association of Pharmaceutical Scientists’s (AAPS) 2026 PharmSci 360 conference, Thomas Luechtefeld, PhD, founder and CEO of Insilica, a US-based scientific artificial intelligence (AI) company, discusses how agentic AI systems are changing computational toxicology workflows and why auditability matters. Dr Luechtefeld is scheduled to present “From Read-Across to Generative AI: The Startup Stack for Animal-Free Toxicology” during the PharmSci 360 Hot Topic session “The Countdown to Post-Animal Drug Development: Disruptors on the Ground” on October 28, 2026.

In addition, auditable systems must show which databases were queried, what results were returned, and whether the process succeeded. Such systems are nearing adoption for real-world decisions and already support research direction, data gap identification, and testing prioritization.

Dr Luechtefeld explains that computational toxicology is accelerating, with the core opportunity being in retrieval and auditable decision making. Systems must retrieve database information comprehensively and reproducibly, organize document information, and incorporate regulator guidance into a decision context, he states, noting that an agent or a human can review relevant information and reach a decision.

Dr Luechtefeld says agentic systems will begin informing risk and hazard decisions and benchmarking alone will not suffice. Users must determine which experiment was performed, what the outcome was, and how far to trust it.

"It's going to be more and more important that people can trace information, trace decision making, and ground everything to truth," Dr Luechtefeld asserts.

Why are single predictive models no longer enough for toxicology decisions?

Dr Luechtefeld points out that the computational models that dominated the past decade are becoming a smaller part of a larger ecosystem. Quantitative structure-activity relationship (QSAR) models and molecular simulation retain value, he explains, but the decision context must broaden. When extensive experimental data exist for a chemical, he adds, those data should inform the decision rather than a QSAR prediction.

In addition, auditable systems must show which databases were queried, what results were returned, and whether the process succeeded, Dr Luechtefeld says. Such systems are nearing adoption for real-world decisions and already support research direction, data gap identification, and testing prioritization, he emphasizes. He expects significant changes within 1 to 2 years in how toxicology teams work.

AAPS 2026 PharmSci 360 will be held October 25-28 in New Orleans. Attendees can visit Insilica at Booth 919.

Explore more of our conference coverage here.

About the speaker

Thomas Luechtefeld, PhD, Founder and CEO, Insilica

Dr Luechtefeld earned his PhD in Environmental Health Sciences from the Johns Hopkins Bloomberg School of Public Health, where he developed RASAR, a machine learning approach to read-across that reached 87% balanced accuracy across nine OECD test types and 190,000 chemicals, exceeding the reproducibility of the animal tests against which the RASAR approach was benchmarked. Dr Luechtefeld’s work is funded by the National Institutes of Health, the National Science Foundation, an FDA research contract, and Horizon Europe, including the ONTOX program for non-animal methods. He recently opened ToxIndex to academic and regulatory researchers at no cost.