Ep199: Martin Burke on Making Small Molecule Medicines for the AI Era

What It Covers

University of Illinois chemistry professor Martin Burke explains how his block chemistry platform—using iterative carbon-carbon bond formation to synthesize small molecules modularly—enables full automation and AI integration. His startup Excelsior Sciences raised $95M to compress drug discovery cycle times from months to weeks, while his academic lab pursues antifungal and iron-transport therapeutics now in Phase 2 trials.

Key Questions Answered

• •Block Chemistry Automation: Traditional synthetic organic chemistry relies on ~1,000 different reactions under thousands of conditions, making automation nearly impossible. Burke's block chemistry uses one repeated carbon-carbon bond-forming reaction with prefabricated building blocks, making the process robot-executable. This compresses lead optimization cycle times from 3–6 months down to 1–2 weeks, fundamentally changing the economics of small molecule drug discovery.
• •AI-Ready Molecular Data: For AI to build predictive foundation models on small molecules, training data must be modular and structurally consistent—analogous to tokens in large language models. Burke's block chemistry generates exactly this format: each molecular building block functions as a token, enabling active learning loops where AI agents select next compounds, robots synthesize them, automated assays test them, and results feed back to the AI within days.
• •Drug Repurposing at the Block Level: Rather than screening random compound libraries, Excelsior Sciences extracts the most common structural blocks from all ~100 FDA-approved kinase inhibitors using Bartosz Grzybowski's algorithm. Those blocks become physical reagents on synthesis robots, generating new candidates built from proven drug-like components. This approach applies across GPCRs, natural product classes, and materials science, providing a validated head start for every new program.
• •Mechanism-First Toxicity Separation: Amphotericin B's kidney toxicity was attributed for decades to its ion channel formation—a conclusion appearing in thousands of papers. Burke's lab synthesized a single-atom derivative removing the hydroxyl group believed critical for channel formation, demonstrating the molecule still killed fungal cells without forming channels. The actual toxicity mechanism is sterol extraction via membrane-surface sponge assembly, enabling rational design of compounds that bind ergosterol in fungi while sparing human cholesterol.
• •Persistence Through Platform Deprioritization: When Revolution Medicines deprioritized the amphotericin antifungal program to focus on KRAS inhibitors, Burke recovered the IP, continued academic development, and identified a next-generation compound series. EL-219, now in Phase 2 trials through Elion Therapeutics, selectively extracts ergosterol from fungal membranes while leaving human cholesterol intact—a result only achievable after obtaining high-resolution structural data of the sterol-sponge complex via solid-state NMR with Chad Rienstra.
• •Iron Mobilization as Therapeutic Target: Ferroportin deficiency—whether genetic or inflammation-induced—causes iron to become trapped, producing anemia affecting millions globally. Graduate student Tony Grillo screened compounds using ferroportin-deficient yeast that cannot grow without an external iron transporter, identifying hinokitiol from Taiwanese cypress bark as a functional replacement. This molecular prosthetic concept, now advanced by Cajal Therapeutics in Seattle, demonstrates that small molecules can restore lost protein transport function rather than merely inhibiting targets.