AI药物开发平台开放公益使用，应对“被商业忽视的疾病”

Core Viewpoint - The Global Health Drug Discovery Institute (GHDDI) has launched an AI-driven drug discovery platform named "AI Kongming," which integrates generative AI and multimodal large model technologies to design potential drug candidates for various diseases, particularly those neglected by commercial interests, such as tuberculosis and malaria [1][3]. Group 1: Platform Features - "AI Kongming" covers the entire drug design process from target structure analysis to AI molecule generation and optimization, active screening, and drug-like property evaluation, aiming to break down traditional barriers in drug development [3]. - The platform is currently available for free, particularly for diseases like malaria and tuberculosis, with a molecular library that is also freely accessible [1][7]. - The platform's "Lian Nu" module can generate new molecules based on existing compound data, significantly improving the efficiency of drug design [4][5]. Group 2: Efficiency Improvements - The "Lian Nu" module has demonstrated a 43% increase in the proportion of active molecules after two iterations for viral targets, and for malaria targets, it achieved 17% and 21% active compounds in the first and second rounds, respectively [5]. - The "Xing Xiang" module enhances the speed of candidate molecule screening, allowing for rapid identification of the most promising candidates from a large pool of generated molecules [5]. - The "Ba Zhen" module quickly assesses the safety and drug-like potential of candidate molecules, integrating nearly 70 pharmacokinetic and toxicological properties for rapid evaluation [6]. Group 3: Database and Results - The "Jin Nang" database provides a wealth of novel compound data for tuberculosis and malaria, generating approximately 3,000 molecules for each of about 400 targets, resulting in a library of millions of potential candidates [6]. - In a comparison of traditional high-throughput screening methods, the "Jin Nang" database achieved a hit rate of approximately 38.5% with only 13 synthesized molecules, significantly reducing the development time by six times [6]. Group 4: Challenges and Future Directions - Current challenges in AI-driven drug discovery include data scarcity, as the pharmaceutical field lacks sufficient labeled data for effective AI training, which limits the potential for further advancements [10]. - There is a call for collaboration across scientific communities to enhance data sharing and improve the overall effectiveness of AI in drug discovery [10].