Shenzhen Key Laboratory of Steroid Drug Discovery and Development, School of Medicine, The Chinese University of Hong Kong, Shenzhen, Guangdong, 518172, China.
Shenzhen Key Laboratory of Steroid Drug Discovery and Development, School of Medicine, The Chinese University of Hong Kong, Shenzhen, Guangdong, 518172, China; Shenzhen Bay Laboratory, Shenzhen, 518055, China.
特异性抗体可以以高亲和力和特异性与蛋白质抗原结合，这一特性使其成为最好的基于蛋白质的治疗药物之一。准确预测抗体-蛋白质抗原结合亲和力对于设计有效的抗体至关重要。当前的蛋白质-蛋白质结合亲和力预测方法通常无法以相当的准确度预测抗体-蛋白质抗原复合物的结合亲和力。在这里，根据抗体-抗原复合物中存在的不同类型的界面和表面积，开发了针对抗体-抗原结合亲和力预测的新模型。基于接触的描述符也用于构建或训练抗体-蛋白质抗原结合亲和力预测的不同模型。本研究结果表明（
i
) 就预测能力而言，基于区域的描述符略好于基于接触的描述符；(
ii
) 抗体-蛋白质抗原结合亲和力预测的新模型优于以前用于预测蛋白质-蛋白质结合亲和力的通用模型；(
iii
) 在这项工作中开发的最好的基于区域和基于接触的模型的性能优于最近开发的基于图形的模型的性能（
即
, CSM-AB) 特异性抗体-蛋白抗原结合亲和力预测。在这项工作中开发的新模型不仅有助于理解抗体-蛋白质抗原相互作用的潜在机制，而且还将在基于抗体的疗法的设计和虚拟筛选中具有一定的实用性。
Specific antibodies can bind to protein antigens with high affinity and specificity, and this property makes them one of the best protein-based therapeutics. Accurate prediction of antibody‒protein antigen binding affinity is crucial for designing effective antibodies. The current predictive methods for protein‒protein binding affinity usually fail to predict the binding affinity of an antibody‒protein antigen complex with a comparable level of accuracy. Here, new models specific for antibody‒antigen binding affinity prediction are developed according to the different types of interface and surface areas present in antibody‒antigen complex. The contacts-based descriptors are also employed to construct or train different models specific for antibody‒protein antigen binding affinity prediction. The results of this study show that (
i
) the area-based descriptors are slightly better than the contacts-based descriptors in terms of the predictive power; (
ii
) the new models specific for antibody‒protein antigen binding affinity prediction are superior to the previously-used general models for predicting the protein‒protein binding affinities; (
iii
) the performances of the best area-based and contacts-based models developed in this work are better than the performances of a recently-developed graph-based model (
i.e.
, CSM-AB) specific for antibody‒protein antigen binding affinity prediction. The new models developed in this work would not only help understand the mechanisms underlying antibody‒protein antigen interactions, but would also be of some applicable utility in the design and virtual screening of antibody-based therapeutics.
