China Debuts AI-Powered GalaxyVS Platform for Drug Discovery

China has taken a major step in AI-driven pharmaceutical research with the launch of GalaxyVS, an artificial intelligence-powered virtual screening platform designed to accelerate the discovery of innovative drugs. Unveiled on May 30 in Tianjin, the platform combines the next-generation Tianhe supercomputing system with the DrugCLIP AI model to screen near-trillion-scale compound libraries at unprecedented speed, according to Xinhua News Agency.

A New Approach to Drug Discovery

Virtual screening is a critical step in drug development, allowing researchers to rapidly identify promising small molecules from massive compound libraries that could target specific disease-related proteins. Traditional experimental screening is costly and time-consuming, while conventional molecular docking methods struggle with computational efficiency as compound libraries approach the trillion-compound scale.

GalaxyVS addresses this bottleneck by transforming the traditional protein-ligand docking process — which requires one-by-one molecular matching — into a massively parallel high-dimensional vector retrieval problem. The system was jointly developed by a team led by Meng Xiangfei, Party Secretary and Chief Scientist of the National Supercomputing Center in Tianjin, and Professor Lan Yanyan from Tsinghua University’s Institute for AI Industry Research (AIR).

“The major breakthrough of GalaxyVS is that it is not a simple scaling up of existing models, but rather a complete reconstruction of a full-process platform around a near-trillion-level chemical space,” Meng said, as reported by Xinhua. The platform integrates AI models, supercomputing, high-performance retrieval, and medicinal chemistry constraints into an end-to-end system.

Built on a Scientific Foundation

At the heart of GalaxyVS lies DrugCLIP, an AI-driven ultra-high-throughput virtual screening platform developed by a Tsinghua University team led by Professor Lan Yanyan. DrugCLIP was published in the journal Science on January 9, 2026, where it demonstrated a million-fold speed improvement over traditional screening methods.

The DrugCLIP model maps protein pockets and small molecules into a unified vector space, enabling ultra-fast semantic retrieval rather than the slow, one-by-one molecular docking of conventional approaches. According to Tsinghua University’s official announcement, a single computing node with 128 CPU cores and 8 GPUs can achieve trillion-scale protein pocket-small molecule scoring throughput per day.

Using DrugCLIP, the team completed the first-ever human genome-scale virtual screening, covering approximately 10,000 protein targets and 20,000 protein pockets, analyzing over 500 million drug-like small molecules. The project enriched more than 2 million potential active molecules, building what is believed to be the largest known protein-ligand screening database, which has been made freely available to the global research community.

Proven Results in the Lab

DrugCLIP’s predictions have been validated experimentally. For the norepinephrine transporter (NET) target — relevant to antidepressant development — 15% of the top-scoring molecules were effective inhibitors, with 12 molecules showing better binding than the existing antidepressant bupropion. For the TRIP12 E3 ubiquitin ligase, a cancer-related target, 10 molecules demonstrated binding ability, with two showing inhibitory activity.

As of January 2026, the screening platform had already served over 1,400 users and completed more than 13,500 screening tasks, indicating strong early adoption by the research community.

Implications for Drug Development

The launch of GalaxyVS marks a significant milestone in China’s AI-driven drug discovery ecosystem. By coupling the DrugCLIP model with the Tianhe supercomputing infrastructure, the platform addresses a critical bottleneck in pharmaceutical R&D — the ability to screen near-trillion-scale compound libraries against thousands of protein targets simultaneously.

This capability could dramatically shorten drug development cycles and reduce costs, particularly for complex diseases where traditional screening is prohibitively expensive. The ability to screen across the entire human genome also opens up possibilities for discovering novel drug targets and first-in-class drugs for cancer, infectious diseases, and rare conditions.

The initiative aligns with China’s broader national strategy to integrate artificial intelligence with life sciences and healthcare, leveraging its world-class supercomputing infrastructure for biomedical innovation.

What’s Next

With GalaxyVS now operational, researchers and pharmaceutical companies will be watching closely to see which therapeutic areas the platform targets first. The DrugCLIP team has indicated plans to collaborate with academic and industry partners to accelerate drug discovery in oncology, infectious diseases, and rare diseases. The platform’s commercial model — whether it will operate on a fee-for-service basis for pharmaceutical companies — remains to be clarified.

As synthesizable compound libraries continue their rapid expansion toward the trillion-compound scale, platforms like GalaxyVS represent a necessary evolution in how the world discovers new medicines.