Making antibody generation rapid, scalable, and democratic through machine learning and continuous evolution

项目摘要/摘要 单抗在生命科学中的重要性怎么强调都不为过。抗体是生物医学中的重要工具 研究和诊断(例如，免疫印迹、免疫沉淀、细胞学、生物标记物发现和组织学) 是增长最快的治疗学类别之一，是癌症治疗中无数新策略的基础，例如 检查点抑制剂，这是革命性的治疗。不幸的是，当前用于生成定制的方法 抗体，包括动物免疫和噬菌体展示，速度慢，成本高，大多数研究人员无法获得，而且通常 不成功。我们提出了自主进化的酵母展示抗体(AEGYS)，这是一个连续和 针对定制抗原的高质量抗体的快速进化，只需要简单地培养酵母细胞。我们 我相信这可以通过结合尖端的生成性机器学习算法来设计抗体库来实现 通过体内持续进化的新技术和我们将设计的酵母抗原提呈细胞。如果 如果成功，AEGYS将通过将单抗转化为单抗，对整个生物医学产生革命性的影响 转化为快速、可扩展和可访问的过程，任何具有标准分子生物学能力的实验室都可以 只需用一种抗原“免疫”酵母细胞的试管，即可按需产生定制抗体。我们期待着 这种抗体产生的民主化也将导致众包抗体序列数据的爆炸性增长 这将训练我们的机器学习算法，为AEGY设计更好的抗体库，开始一个良性循环。我们 我们将使用AEGYS生成一组针对生物胺的亚型和构象特定的纳米体 受体包括那些对乙酰胆碱、肾上腺素、多巴胺和其他神经递质有反应的受体，所以我们可以 了解它们在神经生物学和成瘾中的作用。

Project Summary/Abstract It is hard to overstate the importance of monoclonal antibodies in the life sciences. Antibodies are critical tools in biomedical research and diagnostics (e.g. western blotting, immunoprecipitation, cytometry, biomarker discovery, and histology), are one of the most rapidly growing class of therapeutics, and are the basis for myriad new strategies in cancer therapy, such as checkpoint inhibitors that are revolutionizing treatment. Unfortunately, current methods for the generation of custom antibodies, including animal immunization and phage display, are slow, costly, inaccessible to most researchers, and often unsuccessful. We propose Autonomously EvolvinG Yeast-displayed antibodieS (AEGYS), a system for the continuous and rapid evolution of high-quality antibodies against custom antigens that requires only the simple culturing of yeast cells. We believe this can be achieved by combining cutting-edge generative machine learning algorithms for antibody library design with a new technology for in vivo continuous evolution and a yeast antigen-presenting cell that we will engineer. If successful, AEGYS should have a transformative impact across the whole of biomedicine by turning monoclonal antibody generation into a rapid, scalable, and accessible process where any lab with standard molecular biology capabilities can generate custom antibodies on demand simply by “immunizing” a test tube of yeast cells with an antigen. We anticipate that this democratization of antibody generation will also result in an explosion of crowdsourced antibody sequence data that will train our machine learning algorithms to design better antibody libraries for AEGYS, starting a virtuous cycle. We ourselves will use AEGYS to generate a panel of subtype- and conformation-specific nanobodies against biogenic amine receptors including those that respond to acetylcholine, adrenaline, dopamine, and other neurotransmitters, so that we can understand their role in neurobiology and addiction.!