Department of Chemistry, Indiana University, Bloomington, IN 47405, USA.
Analyst. 2017 May 2;142(9):1525-1535. doi: 10.1039/c6an02697d.
N-Linked glycans, extracted from patient sera and healthy control individuals, are analyzed by Matrix-assisted laser desorption ionization (MALDI) in combination with ion mobility spectrometry (IMS), mass spectrometry (MS) and pattern recognition methods. MALDI-IMS-MS data were collected in duplicate for 58 serum samples obtained from individuals diagnosed with Barrett's esophagus (BE, 14 patients), high-grade dysplasia (HGD, 7 patients), esophageal adenocarcinoma (EAC, 20 patients) and disease-free control (NC, 17 individuals). A combined mobility distribution of 9 N-linked glycans is established for 90 MALDI-IMS-MS spectra (training set) and analyzed using a genetic algorithm for feature selection and classification. Two models for phenotype delineation are subsequently developed and as a result, the four phenotypes (BE, HGD, EAC and NC) are unequivocally differentiated. Next, the two models are tested against 26 blind measurements. Interestingly, these models allowed for the correct phenotype prediction of as many as 20 blinds. Although applied to a limited number of blind samples, this methodology appears promising as a means of discovering molecules from serum that may have capabilities as markers of disease.
从患者血清和健康对照个体中提取 N-连接聚糖，通过基质辅助激光解吸电离（MALDI）与离子淌度谱（IMS）、质谱（MS）和模式识别方法进行分析。对 58 份血清样本的 MALDI-IMS-MS 数据进行了重复采集，这些样本来自被诊断为巴雷特食管（BE，14 例）、高级别异型增生（HGD，7 例）、食管腺癌（EAC，20 例）和无疾病对照（NC，17 例）的个体。对 90 个 MALDI-IMS-MS 谱（训练集）建立了 9 个 N-连接聚糖的组合淌度分布，并使用遗传算法进行特征选择和分类分析。随后，开发了两种用于表型描绘的模型，结果能够明确区分四种表型（BE、HGD、EAC 和 NC）。接下来，这两个模型对 26 个盲测进行了测试。有趣的是，这些模型对多达 20 个盲测的正确表型预测。虽然仅应用于有限数量的盲样，但该方法似乎有希望发现血清中的分子，这些分子可能具有作为疾病标志物的能力。
