Generative Deep Learning toward Antibody Discovery for the Prevention of Food-Borne Illnesses

根据加拿大公共卫生署的统计，在禽肉中最常见的污染物是肠沙门氏菌、空肠弯曲杆菌、**和产气荚膜梭菌。由**S引起的沙门氏菌病。肠道是人类最常见的人畜共患疾病，每年在美国造成130万人食源性疾病和500多人死亡。然而，尽管这种流行和定期爆发，可用于为这些病原体的控制方法提供信息的分子相互作用研究仍处于起步阶段。AbCelex的工作和专长在于开发创新的单域抗体片段（也称为sabs或AbiBodies），用于消除或显着减少人畜共患病原体在禽肉中的定殖。为了找到最适合这项任务的抗体，使用下一代测序平台对抗体库**进行测序，并对选定的抗体进行**各种生化和离体功能活性筛选以选择先导抗体。当前工作流程的一个限制是，这个测试可能需要几个月的时间才能完成；相反，这项工作通过利用最先进的机器学习算法在硅片库中进行设计，该算法可以生成特定于所选目标的各种序列。这个过程将把库生成的时间缩短到2周。考虑到不需要对库进行排序（因为它们将在计算机中生成），而只对它们进行筛选，这也将节省1.5个月的时间。这个Engage项目的资金将**用于建立一个纳米体抑制剂的综合目录，该目录将用于控制安大略省家禽业的细菌负荷，该行业每年的市场价值为7.5亿美元。

According to statistics accumulated by the Public Health Agency of Canada, S.enterica, Campylobacter jejuni,**and Clostridium perfringens are the most prevalent contaminants in poultry meat. Salmonellosis caused by**S.enterica is the most common zoonotic diseases in humans, responsible for 1.3 million human food-borne**illnesses and more than 500 deaths each year in the US. However, despite this prevalence and regular**outbreaks, studies of the molecular interactions that can be used to inform control methods of these pathogens**are in their infancy. AbCelex's work and expertise lies in the development of innovative single-domain**antibody fragments (also known as sdAbs or AbiBodies) utilized for elimination or significant reduction of**poultry meat colonization by zoonotic pathogens. In order to find the optimal antibodies for this task, a library**of antibodies is sequenced using next generation sequencing platforms, and selected antibodies are subjected to**various biochemical and ex vivo functional activity screens to select lead antibodies. A limitation of this**current workflow is that this testing can take months to accomplish; instead, this work sets out to design in**silico libraries by taking advantage of state-of-the-art machine learning algorithms that can generate varied**sequences that are specific to the selected targets. This process would shorten the timeline of library generation**to 2 weeks. Considering that there is no need to sequence the libraries (as they will be generated in silico), but**only screen them, this would also provide a time saving of 1.5 months. This Engage project funding will be**used to build a comprehensive catalogue of nanobody inhibitors that will be used to control bacterial load in**Ontario's poultry industry, which has a thriving annual market value of $750 million.