MindRank’s AI-Driven Weight-Loss Drug Advances to Phase 3 Clinical Trials
A Hangzhou-based biotech start-up, MindRank, has made significant progress in the development of its weight-loss drug, MDR-001. The company has initiated Phase 3 clinical trials for the drug, marking it as China’s first artificial intelligence-assisted Category 1 new drug to reach this stage. This milestone represents a major step forward in the field of pharmaceutical innovation.
MDR-001 is a small molecule GLP-1 receptor agonist designed with the help of AI. These types of drugs mimic natural hormones that regulate blood sugar and appetite. According to Niu Zhangming, the founder and CEO of MindRank, the drug is the first AI-assisted, Category 1 new drug to reach this advanced stage in China. The company aims for approval in the second half of 2028, which would pave the way for a market launch in 2029.
The development process for MDR-001 has taken about four and a half years, significantly shorter than the typical seven to 10 years required to reach this stage. Niu attributed this efficiency to the use of AI, which has effectively reduced overall research and development costs by at least 60 per cent.
AI in Drug Development: A Game Changer
MindRank leverages AI tools to streamline the drug discovery process. Human specialists can specify a target, typically a protein molecule linked to a disease, and then use the company’s proprietary AI tools to rapidly generate potential drugs. This workflow allows researchers to shortlist and select the most promising candidates from an AI-generated pool.
“It’s like overseeing an automated assembly line,” Niu said. This approach not only speeds up the process but also enhances accuracy and efficiency.
In addition to its drug discovery tools, MindRank has developed a specialized biomedical system using open-source large language models (LLMs) integrated with Retrieval-Augmented Generation (RAG) technology. RAG enables LLMs to access real-time information from external documents, improving the accuracy of target research.
According to Niu, this system has helped improve target research accuracy from an industry average of around 85 per cent to over 97 per cent. Target research accuracy measures how effectively viable targets for treating a disease are identified.
AI also plays a crucial role in assessing a drug’s safety and efficacy through advanced predictive models. These models can handle complex calculations that were previously beyond human capability. However, despite these advancements, humans remain an essential part of the process. Many intermediate steps still require manual software operations.
“It still takes a human to tie the entire process together,” Niu noted. Experts continue to steer high-level strategic planning, making critical decisions such as which targets to prioritize and whether to optimize existing compounds or engineer entirely new ones from scratch.
AI in Science: Transforming Life Sciences
AI for Science (AI4S) is revolutionizing life sciences by integrating AI into research. Demis Hassabis and John Jumper of Google DeepMind were awarded the 2024 Nobel Prize for chemistry for Alphafold, which uses AI to predict the structure of proteins.
Last December, US-based Generate:Biomedicines announced plans to launch two global Phase 3 clinical trials for an AI-engineered antibody to treat asthma. Chinese AI4S companies are also gaining significant traction. Baidu-backed start-up BioMap said it had overtaken Alphabet subsidiary Google DeepMind’s AlphaFold in the commercialization of AI foundation models for drug discovery.
Beijing-based start-up DP Technology recently completed a Series C round of funding, raising US$114 million. AI-powered drug discovery firm Insilico Medicine, which has a business model similar to MindRank, listed in Hong Kong last December. MindRank planned to file for an IPO in Hong Kong this year, with the aim of listing in 2027, according to Niu.
Challenges and Future Prospects
Despite the progress, Niu warned that at the current stage AI may not disrupt the life sciences in the same way it has other fields. “The core issue is that the trial-and-error cycle is far too long,” said Niu, lamenting a problem that plagued biopharmaceutical behemoths and start-ups alike, as evidenced by the industry’s protracted validation timelines.
“For AI4S to deliver meaningful impact in life sciences, we still need to go through a longer cycle of testing and evaluation,” Niu said.
