Institute of Computational Biology, Helmholtz Zentrum München-German Research Center for Environmental Health, 85764, Neuherberg, Germany.
Genos Glycoscience Research Laboratory, 10000, Zagreb, Croatia.
Nat Commun. 2017 Nov 14;8(1):1483. doi: 10.1038/s41467-017-01525-0.
Immunoglobulin G (IgG) is a major effector molecule of the human immune response, and aberrations in IgG glycosylation are linked to various diseases. However, the molecular mechanisms underlying protein glycosylation are still poorly understood. We present a data-driven approach to infer reactions in the IgG glycosylation pathway using large-scale mass-spectrometry measurements. Gaussian graphical models are used to construct association networks from four cohorts. We find that glycan pairs with high partial correlations represent enzymatic reactions in the known glycosylation pathway, and then predict new biochemical reactions using a rule-based approach. Validation is performed using data from a GWAS and results from three in vitro experiments. We show that one predicted reaction is enzymatically feasible and that one rejected reaction does not occur in vitro. Moreover, in contrast to previous knowledge, enzymes involved in our predictions colocalize in the Golgi of two cell lines, further confirming the in silico predictions.
免疫球蛋白 G（IgG）是人体免疫反应的主要效应分子，IgG 糖基化的异常与各种疾病有关。然而，蛋白质糖基化的分子机制仍知之甚少。我们提出了一种数据驱动的方法，使用大规模质谱测量来推断 IgG 糖基化途径中的反应。高斯图模型用于从四个队列构建关联网络。我们发现具有高偏相关的聚糖对代表已知糖基化途径中的酶反应，然后使用基于规则的方法预测新的生化反应。使用来自 GWAS 的数据和三项体外实验的结果进行验证。我们表明，一个预测的反应在酶学上是可行的，并且一个被拒绝的反应在体外不会发生。此外，与先前的知识相反，我们预测中涉及的酶在两条细胞系的高尔基体中共定位，进一步证实了这些预测。
