Computational counterselection identifies nonspecific therapeutic biologic candidates.
Computational counterselection identifies nonspecific therapeutic biologic candidates.
复制标题
DOI：
10.1016/j.crmeth.2022.100254
复制
发表时间：
2022-07-18
期刊：
影响因子：
--
通讯作者：
中科院分区：
文献类型：
--
作者：
有效的生物制剂需要高特异性和有限的脱靶结合，但目前基于亲和选择的发现方法并不能保证这些特性。针对脱靶的分子反选择是一种识别非特异性序列的技术，但实验成本高，并且不能消除大部分非特异性序列。在这里，我们引入了计算反选择，这是一个使用机器学习模型从候选生物制剂池中去除非特异性序列的框架。我们使用抗体单靶点亲和选择的测序数据来演示该方法，绕过组合实验。我们通过进行交叉靶点选择和个体结合试验来确定每种方法在保留靶上特异性抗体和识别和消除靶上非特异性抗体方面的性能，从而表明计算反选择优于分子反选择。此外，我们表明，人们可以使用一个通用模型来识别一般的多特异性抗体序列，该模型训练了来自具有广泛序列潜在亲和力的不相关靶点的亲和力数据。计算反选择识别候选池中的非特异性抗体亲和力的ML模型可以训练用于计算反选择抗体文库中的非特异性可以由一般多特异性序列驱动计算反选择可以识别一般多特异性序列生物制剂，如单克隆抗体治疗，通常通过筛选大型随机多样化文库来发现有希望的序列。虽然这些方法对于识别与感兴趣的靶标具有高亲和力的候选物是有效的，但它们需要使用分子反选择来识别非特异性结合，这利用了选择的非预期靶标的组合，并且可能缺乏灵敏度。治疗药物的非特异性结合可能导致药物开发过程中代价高昂的失败和意想不到的不良健康影响。我们试图开发一种计算方法，通过在抗体亲和选择活动的单目标测序数据上训练机器学习模型，在不需要组合实验的情况下，在过程的早期识别非特异性候选抗体。生物制剂要求对靶点具有高特异性，但目前基于亲和选择的发现方法并不能保证这种特性。Saksena等人提出了一种方法，即计算反选择，该方法使用基于单目标亲和力选择实验的高通量数据训练的亲和力机器学习模型来识别非特异性候选者。
Effective biologics require high specificity and limited off-target binding, but these properties are not guaranteed by current affinity-selection-based discovery methods. Molecular counterselection against off targets is a technique for identifying nonspecific sequences but is experimentally costly and can fail to eliminate a large fraction of nonspecific sequences. Here, we introduce computational counterselection, a framework for removing nonspecific sequences from pools of candidate biologics using machine learning models. We demonstrate the method using sequencing data from single-target affinity selection of antibodies, bypassing combinatorial experiments. We show that computational counterselection outperforms molecular counterselection by performing cross-target selection and individual binding assays to determine the performance of each method at retaining on-target, specific antibodies and identifying and eliminating off-target, nonspecific antibodies. Further, we show that one can identify generally polyspecific antibody sequences using a general model trained on affinity data from unrelated targets with potential affinity for a broad range of sequences. Computational counterselection identifies nonspecific antibodies in candidate pool ML models of affinity can be trained for use in computational counterselection Nonspecificity in antibody libraries can be driven by generally polyspecific sequences Computational counterselection can identify generally polyspecific sequences Biologics, such as monoclonal antibody therapeutics, are routinely discovered via screening large, randomly diversified libraries for promising sequences. While these methods are effective for identifying candidates with high affinity for targets of interest, they require the use of molecular counterselection for identifying nonspecific binding, which utilizes combinations of selected unintended targets and can lack sensitivity. The nonspecific binding of therapeutics can lead to costly failure during drug development and unintended adverse health effects. We sought to develop a computational method for identifying nonspecific antibody candidates early in the process without combinatorial experiments by training machine learning models on single-target sequencing data from antibody affinity-selection campaigns. Biologics require high specificity for targets, but current affinity-selection-based discovery methods do not guarantee this property. Saksena et al. present a method, computational counterselection, that identifies nonspecific candidates using machine learning models of affinity trained on high-throughput data from single-target affinity selection experiments.