China and the US Hold AI Protein Design Competitions Concurrently; Yangtze River Cup Expands Wet-Lab Scale to 7,000

The frontier of AI competition is rapidly extending into scientific research. In the field of new drug R&D, where AI4S (AI for Science) is expected to achieve breakthroughs first, two large-scale AI protein design competitions are advancing in parallel on both sides of the Pacific. In the coming months, a large batch of new proteins designed by AI will enter laboratories for real-world testing, among which may lie candidate molecules for future blockbuster drugs. These costly experiments are not only about competition rankings; following the AI "evolution" logic of "AI designs molecules, experiments verify function, data feeds back to models," these two competitions are attempting to drive a sustainable, iteratively improving R&D process and may become landmark practices in the transformation of the scientific research paradigm. It is worth noting that on the Chinese side, a large amount of organizational and resource coordination work for the "Yangtze River Cup" competition is being led by Shanghai. According to the latest confirmation from the organizing committee, although registration has not yet closed, the number of participating teams has exceeded 1,200, far exceeding expectations. Around the competition, domestic universities, research institutions, and enterprises are accelerating their mobilization to explore the path for China's AI4S. Automated laboratory systems can make the AI protein design verification process more efficient. Framework finalized in three days, over a thousand teams respond. The birth of the "Yangtze River Cup" competition was quite dramatic. Last month, U.S. AI giant Anthropic and Swiss protein engineering company Adaptyv announced a jointly organized large-scale protein design challenge. After participating teams use AI tools to design new proteins, they will receive free access to more than 5,000 rigorous biological experiments ("wet experiments") to test whether their designs work. However, the competition excludes Chinese residents and institutions in China. The news quickly drew attention in domestic scientific circles. Just one week later, the "Yangtze River Cup" AI Protein Design Challenge was officially launched in Shanghai, also promising to provide no fewer than 5,000 free wet experiment opportunities and open to research teams worldwide. Surprisingly, for a competition of this scale, the broad framework was finalized in only three days. Zhong Bozitao, secretary of the organizing committee, told reporters that the idea of holding the competition had been brewing in the first half of the year; the launch of the Anthropic competition suddenly accelerated the "Yangtze River Cup." After initial contact with universities, institutes, and enterprises, the feedback Zhong Bozitao and his colleagues received could be described as "a hundred responses to one call," with all parties stating they would "contribute money if they had money and effort if they had effort," and some units that had not been contacted in time also proactively expressed willingness to support. "Everyone felt this was worth doing," Zhong Bozitao said. Soon, Shanghai Jiao Tong University, Shanghai AI Laboratory, Lingang Laboratory, Changping Laboratory, the University of Hong Kong, the Chinese University of Hong Kong, Zhejiang University, Shanghai Institute for Science of Intelligence, and the "Future Starting Point Community," a science and technology innovation incubation platform under Shanghai International Group, were confirmed as co-organizers, and relevant departments of several universities, as well as computing power and scientific research service enterprises, also agreed to participate as co-organizers. Zhong Bozitao introduced that in previous similar competitions at home and abroad, the scale of wet experiments was usually in the hundreds, while these two competitions have increased the scale by an order of magnitude. It should be pointed out that at present, the industry quotation for the synthesis and testing of each protein is several thousand yuan or even higher. Therefore, to send 5,000 designs into the laboratory, plus various competition organization and supporting expenses, the total investment is very considerable. Forces for organizing the competition quickly assembled, and participants followed one after another. At present, the number of teams registered for the "Yangtze River Cup" has exceeded 1,200, far exceeding the organizing committee's expectation of two to three hundred, including many high-level overseas teams, some even from laboratories of Nobel laureates. The picture shows a corner of an automated laboratory system. 5,000 designs handed over to laboratories for judgment. What exactly are these two highly watched competitions competing on? In fact, the competition systems of the two sides are very similar. Both organizers plan to set questions around five health-related targets, allowing participants to use AI tools to design new proteins that can bind to these targets and achieve expected functions, so as to find potential molecules for subsequent drug research. After receiving the protein sequences submitted by participants (equivalent to design blueprints), the organizers will, according to the rules, select no fewer than 5,000 candidate designs and submit them to third-party R&D and experimental service organizations (CROs), which will "follow the blueprint" to prepare protein molecules and test various indicators related to drug development, such as binding ability to the target. Using a large number of wet experiments as competition support is a common choice of the "Yangtze River Cup" and the Anthropic competition, and in fact it is also a response to the industry's widespread demand. The reporter interviewed a Shanghai participating team mainly engaged in AI-assisted macromolecule drug R&D, whose members mostly have computer science backgrounds. The team admitted that the free wet experiments provided by the "Yangtze River Cup" are one of the values they cherish most: "It helps us more clearly understand the boundaries and potential of AI models in drug design." This view is by no means isolated. Zhong Bozitao revealed that "Yangtze River Cup" participants have different styles: some are good at using general large models for design, some specialize in self-developed models, and some are deeply engaged in traditional pharmaceuticals but also hope to take advantage of the AI wave to try connecting their industry experience with AI tools... Whatever type of team, they all look forward to their designs passing the preliminary selection and reaching the wet experiment stage. Previously, due to the high cost of wet experiments, services yet to be improved, and insufficient data sharing within the industry, these teams generally faced the challenge of data scarcity. According to the arrangement, all experimental data will be made public to researchers worldwide. This means that the questions of these two AI protein design competitions, including the "grading analysis," will provide clues and "textbooks" for a new round of R&D. The image shows a computer-simulated three-dimensional protein molecular structure. Design is advancing rapidly, and data has become the bottleneck. On the surface, the "Yangtze River Cup" seems to have a distinctly head-to-head flavor with the other side of the Pacific. But pouring in massive resources is definitely not just about saving face. The two sides coincidentally holding competitions at this time reflects a transformation sweeping the globe. "AI protein design is at an industry inflection point," said Zhu Xiaowen, head of the Shanghai Future Qidian Community, one of the organizers of the "Yangtze River Cup." Over the past few years, AI technologies represented by AlphaFold have significantly advanced protein structure prediction; but now generative AI has gone further, allowing it to skip molecular structure prediction and directly design proteins that do not exist in nature based on function. This "design on demand" is precisely the core direction explored by the two protein competitions. Zhu Xiaowen believes that as this path is gradually opened up, the cycle and cost of new drug R&D are expected to drop significantly, and humanity's boundary in exploring life sciences will also be further broadened. As one of the important initiators of the "Yangtze River Cup," Hong Liang, a professor at Shanghai Jiao Tong University, is an important scholar in China's AI protein design field. The company he founded, Tianwu Technology, ranks among the world's leaders in using AI to predict protein function. Hong Liang told reporters that over the past year, the level of AI protein design has risen rapidly, and some models have even surpassed the capability boundary of AlphaFold. If traditional drug R&D is like finding a needle in a haystack when searching for target molecules, then the new generation of models may become "finding a needle in a pool." The premise is that the models must continue to evolve. Unlike programming and mathematics research, AI's involvement in life sciences and protein design involves the real physical world, and experiments must be used to test whether its designs can work. Hong Liang believes that experiments are both the driving force of current AI4S development and the main bottleneck. Overall, in the field of protein design, experimental validation currently cannot keep up with the progress of AI design capabilities; even existing experimental results have not been fully utilized. In particular, those designs that failed to meet expectations are very high in both absolute number and proportion, but unlike successful cases, they cannot be written into papers and widely disseminated. These dusty data can actually also effectively help models understand which designs do not work, calibrate prediction accuracy, and improve the next round of design. It is precisely because of this pain point that both protein competitions have clearly promised to make all experimental data public, including results that did not meet expectations. According to Fang Xin, head of the biomanufacturing field at Titan Capital, such an arrangement means that both sides regard this competition as an exploration of new scientific research infrastructure in the AI4S era—that is, to further build a large-scale, automated, and standardized wet experiment platform on the existing foundation. In other words, the two competitions are not only a contest of design solutions, but also faintly reveal competition for leadership in AI4S. Future leaders, in addition to having better models, must also extend their capabilities to manufacturing, experiments, data, and other links to form an interconnected closed loop. Gathering forces from multiple parties to explore a Chinese path. In Zhu Xiaowen's view, the "Yangtze River Cup" is not only a top-level global science and technology competition, but also a milestone practice in the deep integration of computational science and life sciences. As scientific research and industrial forces rapidly gather around the competition, the outside world can use the opportunity to observe the advantages and path of China's AI4S. Hong Liang revealed that leading U.S. AI companies have advantages such as general models, computing power, and funding, and are bringing new R&D paradigms into scientific research. At present, OpenAI's attention seems to be on mathematics; in early October it released more than 700 AI-generated research manuscripts in one go, shaking the mathematics community. By contrast, Anthropic seems to focus more on life sciences, and this competition is also its platform for striving for influence in this field. For China, the advantages are mainly reflected in professional experimental service capabilities and industrial support from R&D to production and application. Hong Liang believes that based on China's complete industrial foundation and supply chain system, these advantages will be difficult for other countries to replicate in the short term and can support the transformation of AI protein design. In fact, with the rapid development of AI protein models, global demand for related wet experiments is growing, and a group of domestic CRO companies has ushered in new opportunities. The stock prices of some listed companies have risen sharply since the second half of the year, reflecting capital market expectations for this track. The "Yangtze River Cup" and the thousands of experiments soon to be realized also provide an opportunity for China to fully mobilize this industrial cluster and explore upstream and downstream linkages. It is reported that most of the leading domestic CRO companies have stated their willingness to provide free wet experiment services for the competition, and the organizing committee initially plans to assign experimental tasks to various companies according to different targets. Of course, China is also making every effort to forge AI4S model capabilities. During the "Yangtze River Cup," the organizers will provide multiple domestic AI models and supporting computing power. Zhong Bozitao said that participants can choose the models provided by the organizers or use their own tools and computing power. Overall, the competition not only brings together forces from all sides, but also promotes cross-cooperation among them. As Zhu Xiaowen sees it, the "Yangtze River Cup" is also a new type of incubation mechanism that "connects" scientific research, industry, infrastructure, and capital to create a vibrant soil in which cutting-edge innovation continues to emerge. The reporter has learned the latest news that, due to the continuous increase in support and the further implementation of experimental resources, the "Yangtze River Cup" has decided to expand the scale of wet experiment validation from 5,000 to 7,000 candidate proteins. Let us wait and see how this AI protein design competition will change the industry.