Anthropic团队利用Claude AI设计了一系列蛋白质小结合剂，用于中和蝎毒中的神经毒素。1在64小时内，Claude使用12种设计工具生成了318,484个蛋白质设计，其中211个被选中供实验测试。1
Adaptyv Bio和Twist Bioscience的实验室随后对这些设计进行了验证。1结果表明，17个设计满足结合标准。1其中针对墨西哥蝎毒素Cn2的单毒素设计表现最佳，60个设计中有11个成功结合，成功率达到18%。1最具潜力的两个结合剂Cn2-B1和Cn2-B2对Cn2具有单位纳摩尔级别的亲和力。1竞争实验进一步表明，Claude设计的Cn2结合剂与已报告能中和该毒素的抗体LR竞争相同的表位。1
蝎子蛰伤是一项全球健康问题，每年影响约120万人，导致3250余人死亡，主要发生在医疗资源有限的地区。1该研究展示了AI在加速药物发现中的潜力，特别是在商业投资不足的被忽视疾病领域。1研究团队已向学术界公开发布了全部211个设计及其相关数据。1
Anthropic's Claude artificial intelligence system has successfully designed protein binders targeting scorpion venom neurotoxins, demonstrating AI's potential to accelerate drug discovery for neglected diseases.1 Over a 64-hour period, Claude generated 318,484 protein designs using 12 different design tools, ultimately delivering 211 designs for experimental validation.1
Laboratory testing conducted by Adaptyv Bio and Twist Bioscience confirmed that 17 of the designs met binding criteria.1 Among these, two binders—designated Cn2-B1 and Cn2-B2—demonstrated single-digit nanomolar binding affinity to the Mexican scorpion toxin Cn2, representing particularly potent candidates.1 For single-toxin designs, the success rate reached 18%, with 11 of 60 designs meeting the binding standard.1 Competition experiments revealed that Claude's Cn2 binders compete for the same epitope as the previously reported antibody LR, which is known to neutralize Cn2.1
The work addresses a significant public health challenge, as scorpion stings annually affect approximately 1.2 million people and cause over 3,250 deaths, predominantly in regions with limited medical resources.1 To support broader research efforts, the team has released all 211 designs and associated data to the academic community.1
评论
还没有评论，欢迎留下第一条。