br capture as many features as possible making acquired
capture as many features as possible, making acquired data potentially less reliable or robust as a result. In contrast, the highly reproducible targeted LC-MS/MS method we presented in this study was observed to have a median CV value of 5.1%, producing metabolic markers pre-viously unreported. Further enhancing our data quality and reliability, all targeted metabolites reported in this study were confirmed using pure standard compounds in lieu of applying database searches for LDN-193189 annotation, as is conventionally done in global profiling approaches.
Many studies regarding biomarkers for breast cancer detection have been published [53–58]. For example, cancer antigen (CA) 15-3 and carcinoembryonic antigen (CEA) have already been used clinically ; however, they are mostly indicative of late-stage metastases and exhibit poor classification accuracy (~36–56%) for early-stage BC . Another study analyzing breast cancer tissue samples using GC–MS discovered 13 tumor markers for discrimination between disease and normal subjects  with sensitivity and specificity of roughly 80%. Notably, certain findings of the current study are echoed by previous literature. Hilvo et al.  used LC-MS to investigate lipid metabolism in relation to breast cancer pathogenesis; they found significantly in-creased levels of palmitate in BC patients of all cancer stages, which was also observed in the current study. Another study found 46 urinary biomarkers for breast cancer diagnosis using LC-MS . Interestingly, the researchers discovered significant alterations in tryptophan meta-bolism due mainly to lowered levels of indole and indoleacetate, a trend that was also seen in the current study through analysis of plasma samples.
The application of metabolomics technology in epidemiological and clinical studies is becoming common practice, but few multicenter metabolomics studies to increase reliability have been conducted. This study considers samples from multiple clinical locations, and we sug-gest a method for attenuating the confounding effect of significantly altered metabolite levels due to geographical location of sample col-lection. The 30 metabolites that were not found to be significantly different between controls (presumably ‘housekeeping’ metabolites) were selected for comparison between cancer patients and healthy subjects and for subsequent biomarker selection. Although this study
proposes a solution to the problem of significant metabolite variation inherent in multicenter metabolomics studies, further studies are war-ranted to validate it.
In our current study, we performed univariate analysis of plasma metabolites between BC patients and healthy controls and observed significant alterations in a variety of the metabolites detected. Furthermore, significantly altered plasma metabolites with q < 0.05 and VIP > 1 from the enhanced PLS-DA model were selected for in-clusion in the biomarker panel (Table 2). Additionally, age was in-cluded as a clinical factor to enhance the VIP-based metabolite model. Moreover, our enrichment, pathway, and factor analyses revealed sig-nificant alterations in (1) arginine/proline degradation, (2) fatty acid biosynthesis, and (3) tryptophan metabolism.
Our results indicate proline to be significantly altered in stage I, II, and II breast cancer patients. This finding is in keeping with recent mechanistic studies that have discovered increased proline dehy-drogenase (Prodh) activity to fuel proline catabolism and consequently lead to increased growth of breast cancer cells in 3D culture and in vivo metastasis formation . Consequently, recent efforts have attempted to target proline catabolism, since metastasizing cancer cells rely on by-products of proline degradation to fuel their increased energy need during the colonization of distant organs. As such, targeting proline metabolism does not affect primary cancer growth or non-transformed cells but, rather, impairs metastasis formation in unaffected organs and may lead to better prognosis.
Our results also indicate palmitate to be significantly overexpressed in BC patients at all cancer stages. As stated before, aberrant metabo-lism is a characteristic feature of breast cancer due to the increased energy needs of tumors. Cancer cells often induce a state of lipolysis and/or increased fatty acid synthesis in an effort to meet those needs . This is known as cancer cachexia and has been estimated to ac-count for nearly 20% of all cancer deaths. One key aspect of cancer treatment is the prevention of tumor growth which, in turn, requires a disruption in cancer cells' energy consumption. Accordingly, studies have focused on reducing levels of endogenously produced fatty acids by inhibiting the activity of enzymes responsible for fatty acid bio-synthesis .
The by-products of tryptophan metabolism, indole and indole-3-acetate, were observed to be significantly altered in stage I and stage III BC subjects, respectively. One line of research has focused on the pro-pagation of rapid-growing ‘progressive’ cancers due to a failure of the immune system to maintain control over budding tumors. As a result, there has been increased interest in cancer's ability to escape the human immune response. Recent advances have shown the consumption of tryptophan to be critical in the escape mechanisms of tumors . Mechanistically, cancers have been shown to upregulate the liver en-zyme tryptophan dioxygenase, thereby driving tryptophan consump-tion to produce kynurenine, an endogenous ligand for the aryl hydro-carbon receptor which mediates invasive tumor growth. This affected pathway allows tumors to overcome the human immune response and is, potentially, reflected in the results of our factor and pathway ana-lyses.