MoleculeMind's Xu Jinbo Achieves Ultra-High-Precision Dynamic Protein Design with MoleculeOS Platform

At the 2025 World Artificial Intelligence Conference (WAIC 2025), MoleculeMind, an AI protein design company founded by Jinhui Xu, a pioneer in AI protein folding, showcased a remarkable achievement, injecting a powerful driving force for transformative change into the bioeconomy.

The world's first fully functional AI protein optimization and design platform, MoleculeOS, has been upgraded and launched. It offers superior complex structure prediction capabilities compared to AlphaFold 3, ultra-high precision protein dynamic design capabilities, and dozens of solutions for the biopharmaceutical and biomanufacturing industries."By leveraging globally leading AI technology, biologists without an AI background can also precisely design protein products with specific functions on their computers," said Xu Jinbo.

As an industry-grade AI protein infrastructure platform independently developed by MoleculeMind, MoleculeOS integrates over ten globally leading AI protein prediction, optimization, and design technologies, including the world's first multimodal AI protein foundational large model NewOrigin (Darwin). It also incorporates scientific computational methods such as molecular dynamics and quantum chemistry, providing comprehensive coverage from static observation to dynamic design.

MoleculeOS Architecture

MoleculeMind's algorithm not only achieves the precision of AlphFold 3, but the predicted structures also possess better physical properties. MoleculeOS can more accurately perform complex structure prediction and design tasks such as antigen-antibody and enzyme-substrate complexes. In protein design, MoleculeMind integrates AI with first principles to break through ultra-high precision molecular dynamic structure prediction and protein dynamic design. The accuracy of molecular simulation has significantly improved compared to industry standards, with efficiency increased a millionfold, reaching industrial-grade levels. Several technical achievements, including zero-shot AI enzyme design and peptide design based on molecular surfaces, have been successively published at top international conferences such as ICML, KDD, and MLHC.

Relying solely on foundational general AI protein technologies is insufficient to fully address industrial challenges. Real-world demands in the industry, such as long-acting drug design, precise drug delivery, and improving enzyme catalytic efficiency, often represent highly complex systematic engineering problems that require the integrated application of multiple AI algorithms targeting various objectives. MoleculeOS encapsulates diverse AI algorithms into automated workflows, establishing a series of drug design, enzyme design workflows, and solutions, including conditionally activated antibody design, scFV design, mini-protein design, peptide design, antibody humanization, affinity maturation, protein developability, and other drug R&D technology platforms, as well as enzyme activity design, enzyme stability design, enzyme expression design, and other industrial enzyme optimization platforms. These have already been validated across multiple industrial projects, enabling "one-click access" to custom proteins with specific functionalities tailored for real-world applications like innovative drug design and synthetic biology.

MoleculeOS also features a conversational AI Agent, allowing biologists without an AI background to quickly and accurately design high-value molecules through dialogue with the AI.

MoleculeOS Interface

As the material basis of life, the bioeconomy sector has a huge demand for protein design. In the past, scientists relied on experimental screening in laboratories to design proteins—a process as painstaking and inefficient as finding a needle in a haystack, with extremely low success rates. Consequently, in the field of innovative drug development, the "double ten rule" (ten years and one billion US dollars) has become an increasingly formidable challenge.

Nowadays, with the support of AI, these industry pain points are being gradually resolved. The new method of "AI design + minimal experimental validation" in generative biology has brought tremendous changes to the entire biological field. Compared to the past, where molecular screening could take months or years in the lab and still might not succeed, biologists now only need to focus on input and results, significantly reducing the burden of hands-on operations and improving R&D efficiency. Moreover, high-value molecules designed based on AI further enhance the success rate of R&D. In a multinational pharmaceutical company's long-acting antibody design project, MoleculeMind designed molecules far superior to those achievable by traditional wet-lab methods in a much shorter time.

Currently, MoleculeMind has opened MoleculeOS to the industry and academia."I hope MoleculeOS can become an amplifier of biologists' capabilities, freeing them from tedious laboratory operations and allowing them to focus on strategic planning and decision-making, leading to more precise, safer, and longer-lasting drugs, as well as higher-performance and lower-cost biologics," said Xu Jinbo.