Yang Tan is a PhD student at Shanghai Jiao Tong University and Shanghai Innovation Institute. His research focuses on artificial intelligence for protein science, including protein foundation models, multimodal protein representation, mutation-effect prediction, protein engineering, and scientific AI agents. He has published over 20 papers in AI for biology in venues including NeurIPS, ICLR, and eLife. He developed VenusREM, which ranked first on the ProteinGym mutation-effect prediction leaderboard at the time of publication, and has led the development of the VenusFactory2 protein engineering platform. His open-source models and datasets have received more than 400,000 downloads on Hugging Face. He also serves as a reviewer for venues including Nature Machine Intelligence, ICLR, and ICML.
Our research on AI for protein R&D has progressed from benchmarking to model development, dry–wet experimental validation, and integrated scientific agent platforms. We first established multi-scale benchmarks, such as VenusX, to evaluate protein foundation models across different functional levels. Building on these benchmarks, we developed structure- and evolution-aware models for protein representation and mutation-effect prediction, such as VenusREM, and validated computational predictions through wet-lab experiments. We further integrated data, models, scientific tools, and experimental feedback into protein engineering platforms, such as VenusFactory2, enabling multi-agent workflows toward closed-loop and increasingly autonomous protein discovery and engineering.