Structural bioinformatics software for epitope selection and antibody engineering

描述（由申请人提供）：人类健康极大地受益于单克隆抗体（mAb）的治疗应用，治疗疼痛和破坏性疾病，如类风湿性关节炎和癌症等。然而，mAb开发是一个费力且耗时的过程。从更快的mAb开发中获得的健康益处是显而易见的，因此非常需要工具来指导科学家发现最有希望的抗原靶点-特别是关于B细胞表位（抗体识别的抗原部分）。在这一领域取得进展的关键障碍是在没有已知结构的情况下无法推断蛋白质序列的构象特征，用于预测线性B细胞表位-已知表位中最大，最多样化和药学上有价值的一类。对现有预测方法的普遍批评是它们不准确并且没有解决B细胞表位的构象性质。 DNASTAR建议创建一个软件管道，指导B细胞表位的预测，模拟单克隆抗体与其实验鉴定的抗原之间的动态结构界面，并通过计算机筛选定点突变来设计具有增强结合亲和力的更有效抗体。第一阶段的目标是改善从靶蛋白序列和实验或预测结构预测抗原肽。为了实现这一目标，DNASTAR与单克隆抗体生产，3D结构预测，蛋白质结构和动力学方面的专家建立了合作关系，包括访问他们的实验方法，数据和软件工具。我们的预测模型将受益于三个关键创新：1）上级数据集和对单克隆抗体生产的专业见解，2）引入最先进的3D结构预测来训练我们的表位预测器，以及3）首次在B细胞表位预测中使用基于结构的蛋白质动力学。 在第一阶段结束时，我们将提供一个增强的序列只有B细胞表位预测模型相比，目前的顶级预测方法（目标1）和一个上级序列和结构为基础的表位预测模型，使用三维结构预测和蛋白质动力学（目标2）。在创建这些模型中，我们将考虑蛋白质序列的化学和物理性质以及介导蛋白质-蛋白质相互作用的生物物理学，包括溶剂可及性，氢键，残基灵活性，结合核和分子表面的几何轮廓。拟议的软件管道将建立在Protean 3D，我们新的分子结构和模拟查看器，并将提高广泛的实验科学家的技术能力，以估计关键的抗原结构特性，从蛋白质没有已知的结构-所有在他们的台式计算机上。在实现这些目标后，科学家们将认识到，仅仅使用氨基酸频率或倾向量表来描述B细胞表位已经不够了。 公共卫生相关性：单克隆抗体是诊断和治疗人类疾病的宝贵工具。不幸的是，目前用于鉴定最有希望的免疫原性靶点的实验方法耗时且不完全有效。通过采用将蛋白质序列信息和来自高质量3D结构预测的结构特征结合在我们的桌面计算机软件产品中的新方法，我们建议提高广泛的生命科学家正确预测适用于他们感兴趣的领域的B细胞表位（抗体识别的抗原部分）的能力。这将加速新的单克隆抗体药物的发现，从而改善许多疾病的人类健康。

DESCRIPTION (provided by applicant): Human health has benefited tremendously from the therapeutic application of monoclonal antibodies (mAb), treating painful and devastating diseases such as rheumatoid arthritis and cancer, among others. However, mAb development is a laborious and time consuming process. The health benefits gained from faster mAb development are clear, creating a great need for tools to guide scientists toward discovering the most promising antigenic targets-particularly with regard to B-cell epitopes (the part of an antigen recognized by an antibody). The critical barrier to progress in this domain is the inability to deduce the conformational characteristics of protein sequence in the absence of known structure for predicting linear B-cell epitopes-the largest, most diverse, and pharmaceutically valuable class of known epitopes. The general criticism of existing prediction methods is that they are inaccurate and do not address the conformational nature of B-cell epitopes. DNASTAR proposes to create a software pipeline that guides the prediction of B-cell epitopes, models the dynamic structural interface between a monoclonal antibody and its experimentally identified antigen, and screens in silico site-directed mutations to engineer more potent antibodies with enhanced binding affinity. The Phase I goal is to improve the prediction of antigenic peptides from target protein sequences and experimental or predicted structures. Toward this goal, DNASTAR has established collaborations with experts in monoclonal antibody production, 3D structure prediction, and protein structure and dynamics, including access to their experimental methods, data, and software tools. Our predictive models will benefit from three key innovations: 1) a superior data set and professional insights into monoclonal antibody production, 2) the introduction of state of the art 3D structure prediction for training our epitope predictors, and 3) the first use of structure-based protein dynamics in B-cell epitope prediction. At the conclusion of Phase I, we will deliver an enhanced sequence-only B-cell epitope prediction model when compared to current top prediction methods (Aim 1) and a superior sequence and structure-based epitope prediction model using 3D structure prediction and protein dynamics (Aim 2). In creating these models, we will account for the chemical and physical properties of a protein sequence and the biophysics that mediate protein-protein interactions, including solvent accessibility, hydrogen bonding, residue flexibility, binding nuclei, and geometric contours of the molecular surface. The proposed software pipeline will be built upon Protean 3D, our new molecular structure and simulation viewer, and will elevate the technical capability of a broad range of experimental scientists to estimate key antigenic structural properties from proteins without known structure-all on their desktop computer. Upon achieving these aims, scientists will recognize that it is no longer adequate to describe B-cell epitopes using amino acid frequencies or propensity scales alone. PUBLIC HEALTH RELEVANCE: Monoclonal antibodies are invaluable tools for diagnosing and treating human diseases. Unfortunately, the experimental methods used today to identify the most promising immunogenic targets are time consuming and less than totally effective. By taking the novel approach of incorporating both protein sequence information and structural features derived from high quality 3D structure predictions within our desktop computer software product, we propose to advance the ability of a broad range of life scientists to properly predict B-cell epitopes (the part of an antigen recognized by an antibody) applicable to their area of interest. This will accelerate the discovery of new monoclonal antibody pharmaceuticals, leading to improved human health across many diseases.