Discriminating between patients with endometriosis and healthy women using gas chromatography-mass spectrometry and chemometrics techniques

Chromatographic fingerprinting is a common method for diagnosis of complex and rare diseases [1]. In this regard, gas chromatography –mass spectrometry (GC-MS) is one of the best options for fingerprinting and identification of chemicals in biological samples [2]. On the other hand, due to the complexity of serum sample matrices and lack of selectivity in analytical instruments, multivariate chemometrics methods have been largely applied to extract maximum useful information from chromatographic data [1]. In the present contribution, a chemometrics-based strategy is proposed for GC-MS fingerprint analysis of endometriosis women. In this study, the blood serum of women with endometriosis (referred to Ibn Sina Infertility and Recurrent Abortion Treatment Center, Endometriosis Clinic) whose disease and severity were confirmed by laparoscopic surgery in 18 samples (three samples related to stage 1 and 2 and 15 samples related to stages 3 and 4 of the disease) were prepared. Also, blood serum of healthy women (confirmed by laparoscopic surgery that they do not have endometriosis) was prepared as a control group for 13 samples. Extraction of polar metabolites and derivation of polar metabolites for sample injection into gas chromatography-mass spectrometry were performed. The GC-MS signals were recorded by taking three scans per sample. GC-MS together with pattern-recognition methods such as principal component analysis (PCA), linear discriminant analysis (LDA) and sparse partial least squares Discriminant Analysis (sPLS-DA)were used to discriminate healthy women and women with endometriosis. The results showed excellent discrimination between the healthy women and women with endometriosis. The obtained results were encouraging and showed a promising potential of GC-MS and chemometrics methodes for detecting healthy women and women with endometriosis type. The proposed approach in this work could classify more than 85% of the samples correctly, in each group.