Software for sequencing human antibody proteins from polyclonal immune responses

对外来病原体的自然免疫反应涉及一种复杂的协调， 细胞，包括选择和分泌抗体进入循环的适应性反应。 从病原体感染中恢复的个体保留免疫记忆， 继续传播病原体特异性抗体对于许多传染病，如呼吸道疾病 合胞病毒、埃博拉病毒和痘病毒，对多种病毒具有中和活性的抗体 从人类幸存者身上发现了菌株发现的抗体高度 作为更广泛人群的潜在生物治疗剂， 被自然优化以抵御人类病原体。 从从冠状病毒感染中恢复的人身上发现抗体的努力是 正在进行中，特别是SARS-CoV-2，但受到现有抗体的限制 发现平台。目前的方法需要筛选活性产生的活B细胞 病原体特异性抗体，对细胞死亡敏感，很少在血液中发现。在 血清中抗体蛋白稳定，病原体反应性抗体丰富。 虽然蛋白质是理想的材料，开始，多克隆抗体（pAb）的表征 蛋白质提出了新的挑战。 质谱和分析的最新进展表明， 当足够匹配的B细胞时，候选物可以来自亲和纯化的pAb蛋白质， 提供了遗传抗体库。我们的目标是开发算法来取代 基因抗体库，并从有限的复杂性中重新鉴定抗体候选物 pAb samples.这是通过使用机器对从头肽测序进行改进来实现的 学习和特异性决定区（CDR 3）和抗体的靶向组装 使用de Bruijn graphs。拟议的软件将提供估计克隆 多样性和候选序列，可以合成和测试反应性。此外 解决传染病方面的需要，这一点体现在以下方面的迫切需要尚未得到满足： 阻止COVID-19大流行，该软件还适用于临床和生物医学研究 自身免疫性疾病的需求，以及替代多克隆抗体的商业利益 具有高度可重复的单克隆抗体等效物的试剂。

The natural immune response to foreign pathogens involves a complex coordination of cells, including an adaptive response to select and secrete antibodies into circulation. Individuals who have recovered from a pathogenic infection retain immune memory and continue to circulate pathogen-specific antibodies. For many infectious diseases like respiratory syncytial virus, Ebola, and poxviruses, antibodies with neutralizing activity to multiple viral strains have been discovered from human survivors. The discovered antibodies are highly valuable as potential biologic therapeutics for the broader population, as the antibodies have been naturally optimized to defend against human pathogens. Efforts to discover antibodies from humans recovering from coronavirus infection are underway, SARS-CoV-2 in particular, but are hampered by the limitations of existing antibody discovery platforms. Current approaches require screening for live B cells actively producing pathogen-specific antibodies, which are sensitive to cell death and rarely found in blood. In contrast, antibody protein is stable and pathogen-reactive antibodies are abundant in serum. While protein is the ideal material to start with, characterization of polyclonal antibody (pAb) protein presents new challenges. Recent advances in mass spectrometry and analysis have shown individual antibody candidates can be derived from affinity-purified pAb proteins when a sufficiently matched B-cell genetic antibody repertoire is provided. We aim to develop algorithms to supplant the need of a genetic antibody repertoire, and de novo identify antibody candidates from limited complexity pAb samples. This is achieved by improvements to de novo peptide sequencing using machine learning, and targeted assembly of specificity determining regions (CDR3s) and antibody frameworks using de Bruijn graphs. The proposed software will provide estimates of clonal diversity and candidate sequences that can be synthesized and tested for reactivity. In addition to addressing needs for infectious diseases, as demonstrated with an urgent unmet need to stop the COVID-19 pandemic, the software also applies to clinical and biomedical research needs in autoimmune disease, and commercial interests in replacing polyclonal antibody reagents with highly reproducible monoclonal antibody equivalents.