Shanghai AI Lab Unveils ViraHInter to Boost Antiviral Drug Discovery
Shanghai Artificial Intelligence Lab, in partnership with Fudan University, Ruijin Hospital affiliated to Shanghai Jiao Tong University School of Medicine, and the Shanghai Institute for Viral Research, has introduced ViraHInter, a novel AI predictive model. This release signifies a major leap in antiviral drug development, enabling the prediction of how viruses hijack human proteins without relying on traditional wet-lab experiments.

ViraHInter leverages dual-modal data from both protein sequences and structures. While conventional methods typically examine either amino acid sequences or 3D structures in isolation, ViraHInter uniquely captures both simultaneously with high precision. The model generates full-atom 3D structures of virus-host protein complexes, mapping atomic relationships to facilitate drug design. Additionally, it employs a protein language model to identify conserved patterns that persist despite rapid viral mutations, significantly enhancing prediction accuracy.
Benchmark tests highlight ViraHInter’s exceptional performance. It achieved a prediction accuracy of 0.50 for virus-human protein interactions, outperforming AlphaFold3 by 4.5 times and surpassing other existing methods. The research team applied the model to three influenza subtypes, successfully identifying 33 shared host factors, which underscores its substantial potential in antiviral drug discovery.
Crucially, ViraHInter demonstrates robust adaptability against newly emerging viruses. Even under strict sequence homology controls, the model maintained high performance, confirming its broad applicability in addressing novel pathogens. These findings offer promising new targets and directions for developing treatments against influenza and coronaviruses.
Related article
Newcomer Education AI JiShu Secures Millions, Launches Software-Hardware Strategy
As artificial intelligence continues to transform industries, the educational sector is seeing a breakthrough with the emergence of a promising new player. On April 22, Guangzhou Jishu Artificial Intelligence Technology Co., Ltd. (“Jishu AI”) announc
Beijing accounts for nearly 30% of China’s large model filings, with 225 approved by the Cyberspace Administration, maintaining its lead in AI innovation
As large AI models shift from innovation to regulatory compliance, Beijing is cementing its status as China’s AI innovation hub. The Beijing Municipal Bureau of Economy and Information Technology reports that 225 large models in the city have passed
Sandbar Raises $23M Series A to Advance AI-Powered Note-Taking Ring
Sandbar, a startup founded by former Meta engineers Mina Fahmi and Kirak Hong, gained significant attention last year with the reveal of its Stream ring, a wearable device designed for note-taking. The company has now secured $23 million in Series A
Related Special Topic Recommendations
Comments (0)
0/500
Shanghai Artificial Intelligence Lab, in partnership with Fudan University, Ruijin Hospital affiliated to Shanghai Jiao Tong University School of Medicine, and the Shanghai Institute for Viral Research, has introduced ViraHInter, a novel AI predictive model. This release signifies a major leap in antiviral drug development, enabling the prediction of how viruses hijack human proteins without relying on traditional wet-lab experiments.

ViraHInter leverages dual-modal data from both protein sequences and structures. While conventional methods typically examine either amino acid sequences or 3D structures in isolation, ViraHInter uniquely captures both simultaneously with high precision. The model generates full-atom 3D structures of virus-host protein complexes, mapping atomic relationships to facilitate drug design. Additionally, it employs a protein language model to identify conserved patterns that persist despite rapid viral mutations, significantly enhancing prediction accuracy.
Benchmark tests highlight ViraHInter’s exceptional performance. It achieved a prediction accuracy of 0.50 for virus-human protein interactions, outperforming AlphaFold3 by 4.5 times and surpassing other existing methods. The research team applied the model to three influenza subtypes, successfully identifying 33 shared host factors, which underscores its substantial potential in antiviral drug discovery.
Crucially, ViraHInter demonstrates robust adaptability against newly emerging viruses. Even under strict sequence homology controls, the model maintained high performance, confirming its broad applicability in addressing novel pathogens. These findings offer promising new targets and directions for developing treatments against influenza and coronaviruses.
Newcomer Education AI JiShu Secures Millions, Launches Software-Hardware Strategy
As artificial intelligence continues to transform industries, the educational sector is seeing a breakthrough with the emergence of a promising new player. On April 22, Guangzhou Jishu Artificial Intelligence Technology Co., Ltd. (“Jishu AI”) announc
Beijing accounts for nearly 30% of China’s large model filings, with 225 approved by the Cyberspace Administration, maintaining its lead in AI innovation
As large AI models shift from innovation to regulatory compliance, Beijing is cementing its status as China’s AI innovation hub. The Beijing Municipal Bureau of Economy and Information Technology reports that 225 large models in the city have passed
Sandbar Raises $23M Series A to Advance AI-Powered Note-Taking Ring
Sandbar, a startup founded by former Meta engineers Mina Fahmi and Kirak Hong, gained significant attention last year with the reveal of its Stream ring, a wearable device designed for note-taking. The company has now secured $23 million in Series A





Home






