“If You’re Not Using AI, Start Now”: Retired Chemist Designs Drug Candidate With Chatbots

A case has emerged in which a retired medicinal chemist designed a hypertension drug candidate using only general-purpose chatbots from Google and OpenAI.

According to TechLifeSciences on the 15th (local time), the American Chemical Society (ACS) said that medicinal chemist Liang Guibai presented at ACS Fall 2026 in Chicago how he used ChatGPT and Gemini to design a renin inhibitor. Chemical & Engineering News (C&EN) reported the presentation.

Liang Guibai is a veteran medicinal chemist who co-founded Sheo Pharmaceuticals after retiring from a large pharmaceutical company. His team used Takeda Pharmaceutical’s TAK-272 as a starting point and asked the chatbots questions about the compound’s structure.

The questions concerned the hydrogen-bond donor role of piperidine, whether the morpholine oxygen was necessary for activity, and concerns about metabolic stability. The chatbots synthesized decades of structure-activity relationship (SAR) literature to provide answers.

“It is impossible for a human chemist to remember all that data and derive a clear picture,” Liang Guibai told C&EN. Based on the answers, the team designed a new chemical series and had a contract research organization (CRO) make 200 compounds over several rounds.

The resulting compound, SHEO-054, was reported to perform better than aliskiren in a monkey model of hypertension. Aliskiren is the only renin inhibitor approved by the U.S. Food and Drug Administration (FDA) and was sold by Novartis as Tekturna.

However, these results were self-reported at a conference presentation and have not yet been published in a paper. The compound’s exact structure was not disclosed; only a general scaffold was presented. C&EN noted that superior performance in a single monkey model is a long way from establishing a development candidate.

Liang Guibai said targets such as renin, which have been extensively studied and have a large body of literature, are particularly well suited to AI use. “If you’re not using AI in medicinal chemistry, start right now,” he emphasized.