Data supplementary for NBsTem webserver (http://www.nbscal.online/).

Nanobodies (VHHs) as effective biorecognition elements, have been widely applied in medicine, environment, agriculture and food fields. The thermostability is a crucial factor for its preservation, transportation and application. To address technical challenge from the scarcity of experimental thermostability data for nanobodies, this study innovatively develops two complementary models inferred from an experimental melting temperature Tm and a novel theoretical Qclass associated with conformation stability to realize reliable prediction of the nanobody thermostability by combining extensive molecular dynamics simulation, antibody language model and fused deep learning framework. The Tm prediction model NBsTem_Tm achieves an external test Pearson of 0.83 and MAE of 2.30 °C, significantly outperforming three competitive models reported. The four-classification model NBsTem_Q achieves accuracy of 0.84 for the external test set. With them, thermostability of the INDI database with tens of millions of unexplored nanobodies are first time evaluated, indicating that about 12% nanobodies have high thermostability (Tm higher than 65 °C and simultaneously being Qclass IV). A universal application strategy based on NBsTem model is further proposed to screen nanobodies as desired biorecognition elements. Finally, a user-friendly webserver NBsTem is developed, which can be served as an effective analysis tool for nanobody design and application.纳米抗体（VHHs）作为高效的生物识别元件，已被广泛应用于医药、环境、农业及食品领域。热稳定性是其保存、运输与应用的关键影响因素。针对纳米抗体实验热稳定性数据稀缺所带来的技术难题，本研究创新性地构建了两个互补模型：其一基于实验熔融温度（Tm），其二则关联构象稳定性的新型理论指标Qclass，并结合大规模分子动力学模拟、抗体语言模型与融合深度学习框架，实现纳米抗体热稳定性的可靠预测。其中，Tm预测模型NBsTem_Tm在外部测试集上的皮尔逊相关系数达0.83，平均绝对误差（MAE）为2.30℃，性能显著优于已有报道的3种对比模型。四分类模型NBsTem_Q在外部测试集上的分类准确率达0.84。借助这两个模型，本研究首次对包含数千万条未注释纳米抗体的INDI数据库进行热稳定性评估，结果显示约12%的纳米抗体具备高热稳定性（Tm高于65℃且同时属于Qclass IV类）。本研究进一步提出了基于NBsTem模型的通用应用策略，用于筛选符合需求的纳米抗体生物识别元件。最后，本研究开发了操作便捷的NBsTem网页服务器，可作为纳米抗体设计与应用的高效分析工具。