Total Synthesis Guidance for Chemists

Introduction:

Total synthesis refers to the process of constructing a complex organic molecule often
with biological activity from simpler, commercially available or naturally occurring starting
materials in the fewest steps possible and high yield. Here, two recent synthetic chemistry
papers published in JACS and Nature Communications were analyzed with Large Multimodal
Models (LMMs) so that a chemist can more easily reproduce or make the desired
compound(s) more pure, faster, or less expensive in their own lab.

Procedure:

Each paper was read manually along with corresponding supplementary and peer
review documents. Prompts with keys were created for LMMs that asked specific information
beyond what was provided by authors to potentially increase the yield of pure products in
addition to locating exactly where the information was found that supported its conclusion.
The LMM or Fine-tuned models were tested on three prompts per paper and included
ChatGPT 4o, Organic Chem Scholar, Chemistry Chem, Scholar GPT, and Scholar AI.
Llama 3.1 405B and Nemotron 4 340B large language models were also used to
generate responses on how to make the total synthesis supplementary methods faster or
less expensive than used in original papers, and were judged by a Cohere for AI model.

August 01, 2024

New AI Drug Discovery  

LLM Drug Discovery Applications