A Live Demonstration of Agentic AI in Biochemistry
You’ve finished generating data and are ready to start the analysis. Where do you start? Perhaps you have little experience with stats and even less with programming? How do you organize yourself? What can you do in an hour?
I will give a live demonstration of Claude Code, Anthropic’s agentic AI coding system, using an example challenge: can we predict whether an arbitrary small molecule drug can penetrate the blood-brain barrier (BBB)?
We will watch Claude Code autonomously plan, summarize literature, execute, debug, cross-validate and visualize an end-to-end analysis pipeline in real time using four sub-agents working in parallel: the BLACKSMITH who performs analysis, the BOOKWORM constantly surveying the literature, the ARTIST who expresses the findings, and the ADVERSARY who is forever skeptical.
The Science: Characterizing Molecules that Cross the Blood-Brain Barrier
The BBB is one of the most consequential gatekeepers in drug discovery — a molecule that cannot cross it cannot treat neurological disease, while one that crosses unexpectedly may cause dangerous CNS side effects. Predicting BBB permeability computationally from molecular structure is a high-value target sitting at the intersection of cheminformatics, AI, and neuropharmacology. We use data from Shaker et al. (2021, Bioinformatics), a large set of molecules measured for their capacity to breach the BBB.
The BLACKSMITH will develop a user interface (GUI) to explore these molecules and build a classifier to predict if a compound will traverse the BBB. The BOOKWORM will query external chemical databases to annotate predictions with real-world drug information. The ARTIST will design informative plots and prepare a PowerPoint of the findings. The ADVERSARY will constantly assess correctness, challenge findings and cross-validate.
All orchestrated from a natural language prompt without writing a single line of code — in 50 minutes.
Why You Should Come
Claude Code’s agentic capabilities extend far beyond coding — it can summarize literature, organize project files, perform data analysis, generate reports, and coordinate workflows. My goal is to show how future AI-assisted science may look and leave you with concrete ideas for how these tools can be applied to your research now.
This seminar is designed for researchers at all levels of computational experience: a biochemist who has never opened a terminal, someone curious about AI-assisted research, or a seasoned bioinformatician who wants Claude Code tips.
Event Details
| Date & Time: April 10, 2026 at 10:30 AM |
Location: MSB-384 |
Contact: michael.hallett@uwo.ca