近日，美国伊利诺伊大学Nicholas C. Wu及其课题组发现，一种可解释的语言模型可用于基于精心策划的流感血凝素抗体预测抗体特异性。相关论文于2024年8月19日在线发表在《免疫》杂志上。

研究人员通过挖掘研究出版物和专利，整理了超过5000个流感血凝素（HA）抗体，揭示了HA头部和干部分子抗体之间的许多不同序列特征。研究人员利用该数据集开发了一种轻量级记忆B细胞语言模型（mBLM）用于基于序列的抗体特异性预测。

模型可解释性分析表明，mBLM能够识别HA干部分子抗体的关键序列特征。此外，通过将mBLM应用于未知表位的HA抗体，研究人员发现并实验验证了许多HA干部分子抗体。总体而言，该研究不仅推动了人们对流感病毒抗体反应的分子理解，还提供了一个有价值的资源，便于将深度学习应用于抗体研究。

据了解，尽管经历了几十年的抗体研究，但仅基于抗体序列预测其特异性仍然具有挑战性。主要障碍包括缺乏适当的模型和训练模型所需数据集的不可获得性。

附：英文原文

Title: An explainable language model for antibody specificity prediction using curated influenza hemagglutinin antibodies

Author: Yiquan Wang, Huibin Lv, Qi Wen Teo, Ruipeng Lei, Akshita B. Gopal, Wenhao O. Ouyang, Yuen-Hei Yeung, Timothy J.C. Tan, Danbi Choi, Ivana R. Shen, Xin Chen, Claire S. Graham, Nicholas C. Wu

Issue&Volume: 2024-08-19

Abstract: Despite decades of antibody research, it remains challenging to predict the specificityof an antibody solely based on its sequence. Two major obstacles are the lack of appropriatemodels and the inaccessibility of datasets for model training. In this study, we curated>5,000 influenza hemagglutinin (HA) antibodies by mining research publications andpatents, which revealed many distinct sequence features between antibodies to HA headand stem domains. We then leveraged this dataset to develop a lightweight memory Bcell language model (mBLM) for sequence-based antibody specificity prediction. Modelexplainability analysis showed that mBLM could identify key sequence features of HAstem antibodies. Additionally, by applying mBLM to HA antibodies with unknown epitopes,we discovered and experimentally validated many HA stem antibodies. Overall, thisstudy not only advances our molecular understanding of the antibody response to theinfluenza virus but also provides a valuable resource for applying deep learning toantibody research.

DOI: 10.1016/j.immuni.2024.07.022

Source: https://www.cell.com/immunity/abstract/S1074-7613(24)00371-6