我们提出了一种基于图神经网络的新型机器学习架构STAG(Structural TCR And pMHC binding specificity prediction Graph neural network),利用TCR和pMHC的建模三维结构进行结合特异性预测。实验表明,与其他基于结构和序列的单模态方法相比,STAG在预测TCR-pMHC结合特异性方面表现更优,并在可解释性上实现显著提升,为免疫治疗和...
binding predictionmemory T-cellmachine learningtwo stagesT cells recognize antigens through the interaction of their T cell receptor (TCR) with a peptide-major histocompatibility complex (pMHC) molecule. Following thymic-positive selection, TCRs in peripheral naive T cells are expected to bind MHC ...
也就是说,该方法依赖于TCR-pMHC结合数据和已知的TCR序列特征来预测抗原特异性,然而由于大型数据集通常具有很多复杂性和非特异性背景噪声,因此难以准确验证TCR-pMHC特异性识别[3]。Zhang等人描述了一种名为ICON (IntegrativeCOntext-specific Normalization) 的数据标准化方法,该方法通过消除背景噪声从10xGenomics pMHC结合...
DNA-barcoded pMHC multimers provide a powerful way to simultaneously label T cells recognizing distinct epitopes. This approach was recently used by 10x Genomics to identify and sequence T cells specific for 44 different epitopes38. In this assay, the binding specificity of a T cell was determined...
在这里使用pMHC捕获的独特TCR分子的计数作为AIR抗原结合亲和力的观察指标,遵循DeepTCR使用的策略。使用唯一的UMI来代表每个独特的TCR分子。UMI是一种分子条形码,在测序过程中提供纠错和更高的准确性。这些UMI是用于对样本库中的每个分子进行唯一标记的短序列。这里主要关注TCR AIR抗原结合亲和力的预测。
免疫检查点抑制剂已在多种肿瘤类型中显示出显著的临床疗效,但其受益患者比例依然较低。肿瘤抗原(即新抗原)与人类白细胞抗原(HLA)及T细胞受体(TCR)的结合决定了抗原的呈递和T细胞激活,是免疫治疗反应中的关键因素。传统的免疫原性预测方法通常仅关注抗原与HLA或TCR的单独结合特性,缺乏对这两者结合的综合评估。
肿瘤新生抗原,由每个患者肿瘤中独特的突变产生。由于其极强的肿瘤细胞特异性,是肿瘤靶向免疫治疗的关键靶点。先前已有临床试验(Carreno et al., 2015; Hilf et al., 2019; Keskin et al., 2019; Ott et al., 2017; Sahin et al., 2017)证明个性化肿瘤新抗原疫苗在患者体内的可行性、安全性和免疫原性。另...
Computational prediction of TCR-pMHC binding In recent years, one great advance in neoantigen prediction has shifted from focusing solely on the antigenic peptide to its interaction with T cell receptor (Table 2). De Neuter et al. first used random forest classifiers and discovered both the length...
Conformationally speaking, catch-slip bonds require partial unfolding of the MHC α1α2–α3 and/or TCR Vα–Cα joints and tilting of the TCR–pMHC bonding interface, a prediction consistent with previous results of SMD simulations and single-molecule experiments6. For single-bond dissociation ...
也就是说,该方法依赖于TCR-pMHC结合数据和已知的TCR序列特征来预测抗原特异性,然而由于大型数据集通常具有很多复杂性和非特异性背景噪声,因此难以准确验证TCR-pMHC特异性识别[3]。Zhang等人描述了一种名为ICON (IntegrativeCOntext-specific Normalization) 的数据标准化方法,该方法通过消除背景噪声从10xGenomics pMHC结合...