MindRank’s oral GLP-1 is the most advanced AI-native drug program to date.

      MindRank, a drug developer from Hangzhou focusing on AI, claims its leading candidate progressed from program initiation to Phase III trials in approximately four and a half years with around $23 million in total research expenditure.

      If this holds true, it provides the clearest indication yet to a pressing question in the industry: does AI genuinely alter the economics of drug development, or is it merely a matter of semantics?

      The candidate, MDR-001, belongs to a highly competitive category. It is an oral small-molecule GLP-1 receptor agonist, targeting the obesity and diabetes market, which has transformed the pharmaceutical landscape over recent years. Existing GLP-1 drugs require injections, whereas a small molecule taken in pill form could be cheaper to manufacture and significantly easier to distribute at scale. This potential has prompted every major pharmaceutical company to pursue similar programs.

      In July, the company secured a $52 million Series B funding round led by institutional and healthcare investors, and it operates a platform called Molecule Arts, integrating biological, chemical, computational, and clinical data into a unified research system.

      In addition to its lead program, MindRank has three IND clearances in both China and the US, plus five preclinical candidates identified. Founder and CEO Zhangming Niu articulates the goal as “translating advancements in computation and artificial intelligence into improved medicines for patients,” a common ambition in this field. The $23 million figure is what sets this company apart from others in similar endeavors.

      Traditional estimates for the cost of developing a drug range from hundreds of millions to a couple of billion dollars, although these numbers include failure costs and vary greatly based on who reports them and for what purpose. While one program reaching Phase III for $23 million isn’t directly comparable to those figures, it is significantly lower by any measure.

      However, there are several caveats to consider. Clinical trials in China are notably less expensive than those in the US or Europe, and the $23 million relates to one specific program rather than a wider portfolio. Additionally, Phase III is often where expensive candidates tend to fail. Late-stage trials are where substantial costs begin, as a Phase III study in obesity necessitates a large number of patients over extended periods. The $23 million that moved MDR-001 to this stage will likely not suffice for its entire journey.

      AI is expected to provide advantages earlier in the process. Areas such as candidate selection, property prediction, and refining the chemical space for synthesis are where an effective model can save extensive laboratory time, and this is the phase MindRank claims to have streamlined.

      The competitive landscape is increasingly crowded with significant investment backing various initiatives. Isomorphic Labs, a spinout from DeepMind, is progressing toward trials, ByteDance has entered via Anew Labs, and Anthropic has acquired a biotech AI startup outright.

      What many of these initiatives lack is a candidate in late-stage human trials. With one already, MindRank is further ahead on the singular measure that ultimately resolves the debate, regardless of the contributions of its platform.

      Currently, none of the clinical data has been published or peer-reviewed, and claims regarding development costs and timelines stem from the company’s own statements rather than from formal filings or independent evaluations. This is typical for private biotech firms, and it necessitates a cautious view regarding the figures.

      Over the past two years, the Chinese biotech sector has emerged as a key source of licensed candidates for Western pharmaceutical companies, with numerous substantial deals made for domestically developed assets at lower expenses. A Phase III oral GLP-1 fits this type of asset perfectly.

      Ultimately, the Phase III results will provide a definitive answer that no platform narrative can override. The drug will either prove effective in a large patient population, or it will not, and no degree of computational sophistication upstream can alter that fundamental outcome.