Identification of B cell subsets based on antigen receptor sequences using deep learning

B cell receptors (BCRs) denote antigen specificity, while corresponding cell subsets indicate B cell functionality. Here, we proposed BCR-SORT, a deep learning model that predicts cell subsets from their corresponding BCR sequences. Subsequently, we applied BCR-SORT on BCR repertoire datasets obtained from autoimmune disease patients who have undergone rituximab (RTX) treatment. Peripheral blood samples were obtained at three time points from one patient with pemphigus vulgaris (at diagnosis, one month after the second dose of RTX, and at relapse) and from one patient with myasthenia gravis (before RTX, 1 week after the first dose of RTX, and three months after RTX). Using BCR-SORT, we identified alterations in B cell subpopulations and observed that antibody-secreting cells exhibited resistance to RTX treatment in both MG and PV patients.

B细胞受体（B cell receptors, BCR）可标识抗原特异性，而其对应的细胞亚群则反映B细胞的功能属性。本研究提出了一款名为BCR-SORT的深度学习模型，能够基于对应B细胞受体序列预测细胞亚群。随后，我们将BCR-SORT应用于接受利妥昔单抗（rituximab, RTX）治疗的自身免疫病患者的B细胞受体库数据集。我们从两名患者身上采集了三个时间点的外周血样本：一名寻常型天疱疮（pemphigus vulgaris, PV）患者的样本分别采集于确诊时、第二次RTX给药后1个月以及疾病复发时；另一名重症肌无力（myasthenia gravis, MG）患者的样本分别采集于RTX治疗前、首次RTX给药后1周以及RTX治疗后3个月。借助BCR-SORT，我们鉴定出B细胞亚群的动态变化，并观察到在MG与PV两名患者中，抗体分泌细胞（antibody-secreting cells）均对RTX治疗表现出耐药性。