Bio-Geometry Raises Significant Funding for AI Life Science Model

Bio-Geometry, an AI-native biotechnology company, has successfully closed a strategic funding round totaling several hundred million yuan. The investment was co-led by prominent firms including Shanghai Biomedical Innovation Conversion Fund, CAS Investment, Morningside Venture Capital, and StarVC, with participation from Gaorong Capital and the Index AI Industry Innovation Fund. Index Capital served as the exclusive financial advisor for the transaction.

The newly acquired capital will be strategically deployed to further develop Bio-Geometry's proprietary 'microscopic world model' for life sciences, named GeoFlow, and to advance its in-house drug discovery pipelines. This funding injection underscores the growing investor confidence in the Bio AI sector, a field increasingly recognized for its transformative potential in drug discovery and development. The life sciences industry is witnessing a significant shift towards computational approaches, driven by advancements in artificial intelligence and a growing need to tackle complex biological challenges.

GeoFlow represents a significant leap in understanding and designing biological molecules at the atomic level. The model accurately simulates interactions between biomolecules such as proteins and DNA, enabling the generative AI to create novel molecules not found in nature. This capability moves the field from merely 'understanding life' to actively 'designing life.' Since its initial release in 2024, GeoFlow has undergone three major iterations. GeoFlow V1 achieved parity with leading protein structure prediction tools like AlphaFold 3. By April 2025, GeoFlow V2 expanded its scope beyond structure prediction to include de novo design of antibodies and vaccines. The latest iteration, GeoFlow V3, released in October 2025, focuses on enhancing binding affinity to the nanomolar (nM) level, a critical benchmark for therapeutic efficacy.

Bio-Geometry has demonstrated the power of GeoFlow V3 through numerous real-world applications. In antibody design, the platform has achieved an impressive average hit rate of nearly 20%, requiring the synthesis and validation of fewer than 50 candidate molecules per target. This significantly reduces the time and cost compared to traditional high-throughput screening methods, shortening lead discovery timelines to as little as three weeks. The company has already established over 20 business development collaborations with pharmaceutical companies globally, focusing on areas such as de novo antibody design, multi-objective optimization of lead molecules, and vaccine development.

A notable success story involves the design of highly specific antibodies for tumor immunotherapy. The challenge lay in differentiating between a target antigen and a highly similar 'twin target' on normal cells. GeoFlow's atomic-level modeling capabilities allowed for the direct incorporation of specificity constraints during the design phase, yielding two high-affinity, highly selective antibodies from fewer than 100 sequences. These antibodies precisely bind the target antigen while avoiding the 'twin target,' ensuring enhanced clinical safety from the outset.

Looking ahead, Bio-Geometry is developing GeoFlow V4, which aims to expand modeling capabilities from individual molecular interactions to entire 'molecular systems.' This next-generation model signifies a move towards designing complex biological systems rather than just individual components. The company also has a robust pipeline of dozens of self-developed assets in synthetic biology, with several already advanced to pilot-scale production and others licensed out through technology transfer and revenue-sharing agreements, indicating a strong commercialization strategy.

The company's leadership includes founder Tang Jian, a distinguished AI4S scientist, and Chief Scientific Advisor Yoshua Bengio, a Turing Award laureate. Their pioneering work in AI-driven drug discovery since 2018 has led to significant academic and industrial breakthroughs, including the application of diffusion models for molecular structure generation and contributions to open-source platforms like TorchDrug and TorchProtein.