Shokry Engy, de Oliveira Anselmo Elcana, Avelino Melissa Ameloti Gomes, de Deus Mariana Moreira, Filho Nelson Roberto Antoniosi
Universidade Federal de Goiás (UFG), Campus II, Samambaia, Instituto de Química (IQ) - Laboratório de Métodos de Extração e Separação (LAMES), CEP: 74690-900 Goiânia, GO, Brazil.
Universidade Federal de Goiás (UFG), Campus II, Samambaia, Instituto de Química (IQ), CEP: 74690-970 Goiânia, GO, Brazil.
J Proteomics. 2017 Apr 21;159:92-101. doi: 10.1016/j.jprot.2017.03.005. Epub 2017 Mar 9.
This work combines the advantages of volatile metabolites profiling as a young growing research field with a non-invasive sampling technique using earwax "a neglected body secretion" for detection and monitoring of biomarkers for diabetes mellitus (types 1 and 2). Earwax samples were collected from 26 diabetic patients of both types, analyzed by headspace gas chromatography mass spectrometry and confronted to the volatile earwax composition of 33 healthy individuals. Data mining analysis was conducted using different models to discriminate the healthy individuals from the diabetic patients and to discriminate between both types of diabetes as well. The model with the best discriminating ability was found to be partial least squares discriminant analysis (PLS-DA) after variable selection. The 6 most important biomarkers were ethanol, acetone, methoxyacetone, hydroxyurea, isobutyraldehyde, and acetic acid. The multivariate model constructed was validated using a test data set and was able to correctly predict all the samples. The receiver operating characteristic (ROC) curves were built for the 6 variables for diabetes types 1 and 2 diagnoses. Among the 6 variables selected, methoxyacetone was the only biomarker able solely to perfectly discriminate between diabetes types 1 and 2. The method is simple, non-invasive, accurate, and highly accepted by patients.
Our method involves a volatolomic approach by headspace gas chromatography coupled with mass spectrometry as a single analytical technique combined with multivariate data analysis to detect biomarkers of diabetes in earwax samples. Our method was able to discriminate with high accuracy between 33 healthy controls and 26 diabetic patients as well as its types (1 and 2). Our method employing earwax, a "neglected biological matrix" not only has the advantage of non-invasive sampling but also overcomes the limitations of the applied procedures in other biological samples, involving no or minimum sample pretreatment, no external contamination and utilizing a simple sample collection technique.
这项工作结合了挥发性代谢物分析这一新兴研究领域的优势与一种非侵入性采样技术，该技术使用耳垢（一种被忽视的身体分泌物）来检测和监测糖尿病（1型和2型）的生物标志物。从26名两种类型的糖尿病患者中采集耳垢样本，通过顶空气相色谱 - 质谱联用仪进行分析，并与33名健康个体的挥发性耳垢成分进行对比。使用不同模型进行数据挖掘分析，以区分健康个体与糖尿病患者，并区分两种类型的糖尿病。经过变量选择后，发现具有最佳区分能力的模型是偏最小二乘判别分析（PLS - DA）。6种最重要的生物标志物是乙醇、丙酮、甲氧基丙酮、羟基脲、异丁醛和乙酸。构建的多变量模型使用测试数据集进行验证，能够正确预测所有样本。针对1型和2型糖尿病诊断的6个变量绘制了受试者工作特征（ROC）曲线。在所选的6个变量中，甲氧基丙酮是唯一能够完美区分1型和2型糖尿病的生物标志物。该方法简单、非侵入性、准确且患者接受度高。
我们的方法涉及一种通过顶空气相色谱与质谱联用作为单一分析技术，并结合多变量数据分析来检测耳垢样本中糖尿病生物标志物的挥发组学方法。我们的方法能够高精度地区分33名健康对照者和26名糖尿病患者及其类型（1型和2型）。我们采用耳垢这种“被忽视的生物基质”的方法不仅具有非侵入性采样的优势，还克服了其他生物样本应用程序的局限性，无需或只需最少的样本预处理，无外部污染，并采用简单的样本采集技术。
