MENU

OpenAI and Molecule.one: Improving Chan-Lam Coupling Yields via GPT-5.4

Business + IT Japan
Overview
OpenAI and Molecule.one integrated GPT-5.4 with the Maria experimental platform to improve yields in Chan-Lam coupling reactions, a long-standing bottleneck in pharmaceutical synthesis. The semi-autonomous system executed over 10,000 experiments in three months, demonstrating that AI-driven hypothesis generation and automated validation can significantly accelerate drug discovery.
In Depth

Key Findings

The integration of GPT-5.4 and the Maria platform successfully increased the average yield of Chan-Lam coupling reactions from 16.6% to 25.2%. Specifically, the proportion of reactions achieving yields over 30% rose from 15.6% to 37.5%. Human verification in flask environments confirmed yield improvements in 11 out of 14 substrate pairs, with 8 pairs showing more than double the yield compared to control conditions without additives.

Technical Details

The project targeted the formation of carbon-nitrogen bonds, specifically focusing on 1-primary sulfonamides, which historically exhibit low yields. The workflow functioned as a semi-autonomous loop:

  • Hypothesis Generation: GPT-5.4 analyzed academic literature to identify priority substrates and automatically generated thousands of improvement proposals.
  • Human Intervention: Human chemists screened the AI’s proposals and applied minor corrections, such as excluding specific solvents.
  • Automated Execution: The Maria platform, consisting of specialized models and an automated laboratory, executed 10,080 microliter-scale reactions across two cycles.

The AI identified that adding TEMPO, a mild oxidant, improved the reaction. This was validated across 88% of tested boronic acids and 83% of sulfonamides.

Background and Context

Chan-Lam coupling is critical for synthesizing various pharmaceuticals, including anticancer, antibacterial, and antihypertensive drugs. Traditionally, the low efficiency of this reaction in the early stages of drug discovery has acted as a significant bottleneck. By automating the path from literature review to data analysis, the team completed the process from project start to expert evaluation in approximately three months.

Significance and Outlook

This demonstration highlights the potential of “AI scientists” to handle physical experiments autonomously. The ability to compress long-term verification processes into a few months suggests a structural shift in process chemistry and industrial scientific research.

Source: https://news.yahoo.co.jp/articles/497f70451ca7ac00516046d5c810f102340faec8

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.

  • Your email and selected fields are used only to deliver the newsletter.
  • We never share your information with third parties.
  • You can unsubscribe anytime via the link in each email.

See our Privacy Policy for details.

Takes about a minute · Unsubscribe anytime

Published by Troy-Technical, an independent site run by one engineer with a career in materials development.
About the author / Contact info@troy-technical.jp
Let's share this post !

Author of this article

TOC