A groundbreaking advancement in drug discovery has emerged from the Institute for AI Industry Research (AIR) at Tsinghua University, led by Professor Lan Yanyan, in collaboration with researchers from the School of Life Sciences and the Department of Chemistry. The team has developed DrugCLIP, an AI-powered ultra-high-throughput virtual screening platform that accelerates drug discovery by up to a million times compared to traditional methods. Currently, only about 10% of the human genome’s potential drug targets have been explored. With over 20,000 proteins encoded in the human genome, the challenge lies in efficiently identifying promising small-molecule compounds from an astronomically vast chemical space. Conventional molecular docking tools face severe limitations—screening 10,000 protein targets against 109 candidate molecules would require approximately 10¹³ protein-ligand scoring calculations. Even on a single computer, this would take hundreds of years, making large-scale exploration impractical. DrugCLIP overcomes this bottleneck by transforming the traditional docking process into a semantic retrieval task in vector space. Using just a single computing node equipped with 128-core CPUs and 8 GPUs, the platform achieves a daily throughput of trillions of protein-pocket and small-molecule pair evaluations—dramatically reducing computation time from centuries to a single day. This innovation builds on the success of AlphaFold, which solved the protein structure prediction problem. DrugCLIP now bridges the gap from structure prediction to drug discovery, enabling genome-scale virtual screening for the first time. The platform’s accuracy was validated early on through wet-lab experiments. In collaboration with Professor Yan Chuangye’s team, DrugCLIP screened 1.6 million candidate molecules for the norepinephrine transporter (NET), a key target for depression. The model identified around 100 high-scoring molecules, and experimental testing confirmed that 15% were effective inhibitors—12 of which showed stronger binding than the existing antidepressant bupropion. Further validation came from cryo-electron microscopy, which resolved the structures of several drug-target complexes, confirming the biological relevance of the predictions. In another study, Professor Liu Lei’s team applied DrugCLIP to TRIP12, an E3 ubiquitin ligase linked to cancer and Parkinson’s disease. Since TRIP12 lacks known ligands or complex structures, the team used its AlphaFold-predicted structure. DrugCLIP screened 1.6 million molecules and identified 50 high-scoring candidates. Experimental results confirmed 10 binders, with two showing significant inhibition of TRIP12’s enzymatic activity—demonstrating the platform’s ability to work effectively with predicted structures and unbound pockets. The research has resulted in the creation of the largest known protein-ligand screening database to date, covering approximately 10,000 protein targets and 20,000 protein pockets, analyzing over 500 million drug-like molecules and identifying more than 2 million potential active compounds. This database is now freely available to the global scientific community to accelerate basic research and early-stage drug discovery. A user-friendly online platform has also been launched, allowing researchers to upload their own protein targets or pockets for customized virtual screening. Within six months of its launch, the platform has served over 1,400 users and completed more than 13,500 screening tasks. The findings were published in Science on January 9 under the title “Deep contrastive learning enables genome-wide virtual screening.” The study was supported by key national programs including the Ministry of Science and Technology’s Key R&D Program, the National Natural Science Foundation of China, and the New Cornerstone Research Fund, as well as contributions from Tsinghua University Wuxi Institute for Applied Technology, Beijing Academy of Artificial Intelligence, and Beijing Structural High-Tech Center. The co-first authors are Jia Yinjun, Gao Bowen, Tan Jiaxin, Zheng Jiqing, and Hong Xin from AIR. The corresponding authors are Professor Lan Yanyan from AIR, along with Associate Professors Zhang Wei and Yan ChuanYe from the School of Life Sciences, and Professor Liu Lei from the Department of Chemistry. The research marks a transformative step toward a more intelligent, efficient, and accessible global drug discovery ecosystem, with future applications focused on cancer, infectious diseases, and rare disorders. The team plans to deepen collaborations across academia and industry to accelerate the development of first-in-class therapeutics.