br There are some limitations in this study This study
There are some limitations in this study. This study considers samples from multiple clinical locations which, although critically im-portant, is underused in metabolomics study design. However, knowl-edge of sample demographics is a limitation. Although statistically important clinical features such as age were known, we failed to collect other demographic information such as body mass index or smoking and drinking status, which may be confounding factors for the asso-ciations between metabolite levels and breast cancer status.
Although the analytical platform was optimized for the detection of over 400 metabolites, 30 potential markers were included in further data analysis when accounting for reliability and non-significance be-tween control samples collected from two clinical sites. Results of our representation analysis indicate that the metabolic profile generated by these 30 metabolites is reflective of 27 metabolic pathways (see Supplemental Fig. S2). Likewise, previous studies have used 20 com-pounds to infer tryptophan-induced pathogenesis of breast cancer , and 25 metabolites for discrimination between lung cancer patients and age-matched controls . Also, we performed pathway-based analyses to observe higher-order effects due to breast cancer . It should also be noted that there is a great need for breast cancer metabolomics studies utilizing multi-center designs ; our study aims to address that need while also offering a simple way to account for possible metabolic variations in samples taken from different clinical locations.
Furthermore, more research is needed to study 1062368-24-4 related to molecular subtypes of breast cancer. In this study, no significant difference in metabolites was observed between ER/PR+, HER2+ vs. ER/PR+, HER2−, and triple negative vs. non-triple negative patients. However, 15 metabolites had p < 0.05 when comparing ER, PR, and HER2 positive and negative patients, as well as staged BC patients, al-though none of these 15 metabolites remained significant after FDR-correction. Future studies should also examine larger cohorts from multiple locations to further validate the altered metabolites and me-tabolic pathways related to BC pathogenesis discovered in our study [66,67]. To the best of our knowledge, this is the first study using an LC-MS/MS metabolomics approach to analyze plasma samples from two clinical sites for breast cancer diagnosis. Our predictive model de-monstrates relatively good performance (89%) for detection of stage I and II breast cancer with specificity of 75% when sensitivity is 80%. This study provides a strong rationale for the development of larger multi-site projects to validate the findings across population groups and further advance the development of accurate tools for clinical risk prediction of breast cancer.
This study is part of a growing body of literature in which an LC-MS/MS targeted plasma metabolic profiling approach has been applied for the comparison of BC patients and healthy controls. To the best of our knowledge, this is the first targeted approach for breast cancer diagnosis to consider samples from multiple locations. Our results de-monstrate a panel of 18 metabolites with FDR q-value < 0.05 and 6 Journal of Chromatography B 1105 (2019) 26–37
metabolites with both q < 0.05 and VIP > 1. The 6 differential me-tabolites were also shown to be effective for the detection of early stage, localized disease (stage I and II). Application of bioinformatic methods showed underlying disturbances in metabolic pathways related to tumor growth, metastasis, and immune escape mechanisms. Accounting for age, this metabolic profiling method can potentially provide a novel disease biomarker panel for breast cancer.
The research team would like to thank the Fred Hutchinson Cancer Research Center Breast Specimen Repository for their allocation of clinical samples. Support from the College of Health Solutions at Arizona State University and Friends for an Earlier Breast Cancer Test (http://www.earlier.org/) is gratefully acknowledged.
Conflict of interest disclosure
Dr. Daniel Raftery serves as Chief Scientific Officer for and holds equity in Matrix-Bio, Inc. The other authors declare no potential con-flicts of interest.
Appendix A. Supplementary data
M. Fernandez, C. de la Torre Cabrera, C. Ramírez-Tortosa, S. Granados-Principal,
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/neucom
Breast cancer diagnosis through active learning in content-based image retrieval