Workflow
Foldseek
icon
Search documents
诺奖得主David Baker最新论文:AI设计蛋白新突破,精准设计蛋白结合剂,克服“不可成药”靶点
生物世界· 2026-01-27 08:00
Core Insights - The article highlights a significant breakthrough in protein design using conditional RFdiffusion to create high-affinity binding proteins for hydrophilic targets, led by Nobel laureate David Baker [4][7]. Design Strategy - The design strategy involves generating extended beta-sheet structures that geometrically match the edges of the target protein's beta strands through conditional RFdiffusion [5]. - Specially designed hydrogen bond groups are created to complement the polar groups on the target protein [6]. Experimental Validation - This technology overcomes traditional limitations in computational protein design, significantly expanding the range of target proteins for designed binding agents, particularly addressing challenges related to hydrophilic interactions. This advancement holds substantial value for drug development and protein function research [7]. - The designed protein binding agents exhibit high specificity and affinity, achieving picomolar to nanomolar levels of binding affinity for important protein targets such as KIT and PDGFRα [9]. Training and Courses - A series of online courses are offered, including AI protein design, antimicrobial peptide design, and computer-aided drug design, aimed at equipping participants with cutting-edge knowledge and practical skills in protein design [8]. - Various promotional offers are available for course registrations, including discounts for early sign-ups and bundled course registrations [8]. Future Trends - The article emphasizes the importance of AI protein design as a key technology to watch in 2026, with a growing demand for training and resources in this field, as evidenced by the high attendance and positive feedback from previous training sessions [7].
Nature Biotechnology:西湖大学原发杰/常兴团队等开发ProTrek,以自然语言“导航”蛋白质宇宙
生物世界· 2025-10-03 01:00
Core Insights - The article discusses the development of ProTrek, a novel trimodal protein language model that integrates amino acid sequences, three-dimensional structures, and natural language descriptions for advanced protein searches [3][9][20]. Group 1: Challenges and Opportunities in Protein Research - Proteins are essential for life, and understanding the complex relationship between their sequences, structures, and functions is crucial for molecular science and pharmacology [6]. - Traditional tools like BLAST and Foldseek are limited to single-modal comparisons, hindering the discovery of cross-modal relationships between sequences, structures, and functions [6][9]. - Approximately 30% of proteins in the UniProt database remain functionally unannotated due to their distant phylogenetic relationships with known homologs, likened to "dark matter" in the protein universe [6][9]. Group 2: ProTrek's Innovative Framework - ProTrek employs a unique trimodal framework that unifies three core protein information types: amino acid sequences (1D), three-dimensional structures (spatial), and natural language function descriptions (semantic) [9][20]. - The model utilizes a bidirectional alignment framework to establish strong correlations across sequence-structure, structure-function, and function-sequence dimensions [9][21]. Group 3: Performance and Experimental Validation - ProTrek demonstrates superior performance, outperforming existing top methods like ProteinDT and ProtST by over 30-60 times in standard protein function retrieval benchmarks [11][21]. - The model's global representation learning capability allows it to identify convergent evolution proteins that have significant sequence and structure differences but perform similar functions [11][21]. - Experimental validation showed ProTrek's ability to discover new proteins with similar functions to human UDG, achieving higher editing efficiency and lower off-target effects compared to existing tools [15][23]. Group 4: Implications and Future Prospects - ProTrek enhances the efficiency and depth of protein research, facilitating large-scale annotation of unknown protein functions and accelerating enzyme discovery and drug design [18][23]. - The model's integration of complex molecular data with intuitive natural language promotes a better understanding of the protein world [18][23]. - ProTrek's capabilities are expected to lead to new scientific discoveries across various fields of protein science [18][23].