Lingli Deng

Colorectal cancer detection using targeted serum metabolic profiling.

Colorectal cancer (CRC) is one of the most prevalent and deadly cancers in the world. Despite an expanding knowledge of its molecular pathogenesis during the past two decades, robust biomarkers to enable screening, surveillance, and therapy monitoring of CRC are still lacking. In this study, we present a targeted liquid chromatography-tandem mass spectrometry-based metabolic profiling approach for identifying biomarker candidates that could enable highly sensitive and specific CRC detection using human serum samples. In this targeted approach, 158 metabolites from 25 metabolic pathways of potential significance were monitored in 234 serum samples from three groups of patients (66 CRC patients, 76 polyp patients, and 92 healthy controls). Partial least-squares-discriminant analysis (PLS-DA) models were established, which proved to be powerful for distinguishing CRC patients from both healthy controls and polyp patients. Receiver operating characteristic curves generated based on these PLS-DA models showed high sensitivities (0.96 and 0.89, respectively, for differentiating CRC patients from healthy controls or polyp patients), good specificities (0.80 and 0.88), and excellent areas under the curve (0.93 and 0.95). Monte Carlo cross validation was also applied, demonstrating the robust diagnostic power of this metabolic profiling approach.

Targeted serum metabolite profiling and sequential metabolite ratio analysis for colorectal cancer progression monitoring

Colorectal cancer (CRC) is one of the most prevalent cancers worldwide and a major cause of human morbidity and mortality. In addition to early detection, close monitoring of disease progression in CRC can be critical for patient prognosis and treatment decisions. Efforts have been made to develop new methods for improved early detection and patient monitoring; however, research focused on CRC surveillance for treatment response and disease recurrence using metabolomics has yet to be reported. In this proof of concept study, we applied a targeted liquid chromatography tandem mass spectrometry (LC-MS/MS) metabolic profiling approach focused on sequential metabolite ratio analysis of serial serum samples to monitor disease progression from 20 CRC patients. The use of serial samples reduces patient to patient metabolic variability. A partial least squares-discriminant analysis (PLS-DA) model using a panel of five metabolites (succinate, N2, N2-dimethylguanosine, adenine, citraconic acid, and 1-methylguanosine) was established, and excellent model performance (sensitivity = 0.83, specificity = 0.94, area under the receiver operator characteristic curve (AUROC) = 0.91 was obtained, which is superior to the traditional CRC monitoring marker carcinoembryonic antigen (sensitivity = 0.75, specificity = 0.76, AUROC = 0.80). Monte Carlo cross validation was applied, and the robustness of our model was clearly observed by the separation of true classification models from the random permutation models. Our results suggest the potential utility of metabolic profiling for CRC disease monitoring.

Combining NMR and LC/MS Using Backward Variable Elimination: Metabolomics Analysis of Colorectal Cancer, Polyps, and Healthy Controls

Both nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) play important roles in metabolomics. The complementary features of NMR and MS make their combination very attractive; however, currently the vast majority of metabolomics studies use either NMR or MS separately, and variable selection that combines NMR and MS for biomarker identification and statistical modeling is still not well developed. In this study focused on methodology, we developed a backward variable elimination partial least-squares discriminant analysis algorithm embedded with Monte Carlo cross validation (MCCV-BVE-PLSDA), to combine NMR and targeted liquid chromatography (LC)/MS data. Using the metabolomics analysis of serum for the detection of colorectal cancer (CRC) and polyps as an example, we demonstrate that variable selection is vitally important in combining NMR and MS data. The combined approach was better than using NMR or LC/MS data alone in providing significantly improved predictive accuracy in all the pairwise comparisons among CRC, polyps, and healthy controls. Using this approach, we selected a subset of metabolites responsible for the improved separation for each pairwise comparison, and we achieved a comprehensive profile of altered metabolite levels, including those in glycolysis, the TCA cycle, amino acid metabolism, and other pathways that were related to CRC and polyps. MCCV-BVE-PLSDA is straightforward, easy to implement, and highly useful for studying the contribution of each individual variable to multivariate statistical models. On the basis of these results, we recommend using an appropriate variable selection step, such as MCCV-BVE-PLSDA, when analyzing data from multiple analytical platforms to obtain improved statistical performance and a more accurate biological interpretation, especially for biomarker discovery. Importantly, the approach described here is relatively universal and can be easily expanded for combination with other analytical technologies.

Exploring Metabolic Profile Differences between Colorectal Polyp Patients and Controls Using Seemingly Unrelated Regression

Despite the fact that colorectal cancer (CRC) is one of the most prevalent and deadly cancers in the world, the development of improved and robust biomarkers to enable screening, surveillance, and therapy monitoring of CRC continues to be evasive. In particular, patients with colon polyps are at higher risk of developing colon cancer; however, noninvasive methods to identify these patients suffer from poor performance. In consideration of the challenges involved in identifying metabolite biomarkers in individuals with high risk for colon cancer, we have investigated NMR-based metabolite profiling in combination with numerous demographic parameters to investigate the ability of serum metabolites to differentiate polyp patients from healthy subjects. We also investigated the effect of disease risk on different groups of biologically related metabolites. A powerful statistical approach, seemingly unrelated regression (SUR), was used to model the correlated levels of metabolites in the same biological group. The metabolites were found to be significantly affected by demographic covariates such as gender, BMI, BMI(2), and smoking status. After accounting for the effects of the confounding factors, we then investigated potential of metabolites from serum to differentiate patients with polyps and age matched healthy controls. Our results showed that while only valine was slightly associated, individually, with polyp patients, a number of biologically related groups of metabolites were significantly associated with polyps. These results may explain some of the challenges and promise a novel avenue for future metabolite profiling methodologies.

Targeted metabolic profiling of wounds in diabetic and nondiabetic mice.

While cellular metabolism is known to regulate a number of key biological processes such as cell growth and proliferation, its role in wound healing is unknown. We hypothesized that cutaneous injury would induce significant metabolic changes and that the impaired wound healing seen in diabetes would be associated with a dysfunctional metabolic response to injury. We used a targeted metabolomics approach to characterize the metabolic profile of uninjured skin and full‐thickness wounds at day 7 postinjury in nondiabetic (db/‐) and diabetic (db/db) mice. By liquid chromatography mass spectrometry, we identified 129 metabolites among all tissue samples. Principal component analysis demonstrated that uninjured skin and wounds have distinct metabolic profiles and that diabetes alters the metabolic profile of both uninjured skin and wounds. Examining individual metabolites, we identified 62 with a significantly altered response to injury in the diabetic mice, with many of these, including glycine, kynurenate, and OH‐phenylpyruvate, implicated in wound healing for the first time. Thus, we report the first comprehensive analysis of wound metabolic profiles, and our results highlight the potential for metabolomics to identify novel biomarkers and therapeutic targets for improved wound healing outcomes.

NMR-based Metabolite Profiling of Pancreatic Cancer

Metabolite profiles of serum from pancreatic cancer (PC) patients (n=51) and non-disease controls (n=47) were measured using H nuclear magnetic resonance (NMR) spectroscopy with a focus on the metabolic changes associated with PC pathology and the development and external validation of the statistical models developed using the metabolite data. Univariate statistical analysis indicated 42 metabolite features showing significant differences between PC and controls (p<0.05). Based on multivariate regression analysis of the data from 38 PC patients and 32 controls, twelve distinguishing metabolites (alanine, choline, citrate, creatinine, glucose, glutamine, glutamic acid, 3hydroxybutyrate, lactate, lipids, methionine and valine) were determined based on their ranked importance in a partial least square discriminant analysis (PLS-DA) model. A cross-validated regression PLS-DA model built using these metabolites differentiated the cancer and control groups with high accuracy and an area under the receiver operating characteristic curve (AUROC) of 0.95. Notably, external validation of this model using the NMR data from a second, distinct set of samples (13 PC patients and 15 controls) collected approximately 1 year later showed an AUROC of 0.86, which represents very good performance compared to the current approaches for identifying pancreatic cancer patients. Metabolic changes in pancreatic cancer patients compared to healthy controls as shown in this study demonstrate the potential for the development of regression models based on blood metabolites to identify patients with pancreatic cancer.

Altered metabolite levels and correlations in patients with colorectal cancer and polyps detected using seemingly unrelated regression analysis

IntroductionMetabolomics technologies enable the identification of putative biomarkers for numerous diseases; however, the influence of confounding factors on metabolite levels poses a major challenge in moving forward with such metabolites for pre-clinical or clinical applications.ObjectivesTo address this challenge, we analyzed metabolomics data from a colorectal cancer (CRC) study, and used seemingly unrelated regression (SUR) to account for the effects of confounding factors including gender, BMI, age, alcohol use, and smoking.MethodsA SUR model based on 113 serum metabolites quantified using targeted mass spectrometry, identified 20 metabolites that differentiated CRC patients (n = 36), patients with polyp (n = 39), and healthy subjects (n = 83). Models built using different groups of biologically related metabolites achieved improved differentiation and were significant for 26 out of 29 groups. Furthermore, the networks of correlated metabolites constructed for all groups of metabolites using the ParCorA algorithm, before or after application of the SUR model, showed significant alterations for CRC and polyp patients relative to healthy controls.ResultsThe results showed that demographic covariates, such as gender, BMI, BMI2, and smoking status, exhibit significant confounding effects on metabolite levels, which can be modeled effectively.ConclusionThese results not only provide new insights into addressing the major issue of confounding effects in metabolomics analysis, but also shed light on issues related to establishing reliable biomarkers and the biological connections between them in a complex disease.

Abstract B50: Detecting colorectal cancer and polyps using nuclear magnetic resonance spectroscopy and mass spectrometry based metabolomics

Background: Colorectal cancer (CRC) represents the third most prevalent cancer in the US, despite being one of the most preventable. Given the strong connection between CRC and metabolism, a number of these efforts have been made using metabolomics. Currently, the majority of metabolomics studies use either nuclear magnetic resonance (NMR) or mass spectrometry (MS) separately. However, thus far the potential of combining NMR and MS for biomarker discovery and statistical modeling is still poorly recognized and not well developed (such as the challenges of variable selection in the combined data sets). In this study we examine the potential of combining NMR and MS metabolomics for the detection of patients with either CRC or polyps. Methods: A total of 127 serum samples from three groups of age, gender, and BMI-matched subjects (CRC (N=28), polyp patients (N=44) and healthy controls (N=55)) were analyzed, and potential biomarkers were selected from backward variable elimination incorporated multi-block partial least squares-discriminant analysis (BVE-PLSDA). In each iteration, one variable was dropped out, and the remaining variables were used for PLS-DA. The variables with the highest prediction accuracy for the test samples in Monte Carlo cross validation (MCCV) were kept for the next iteration. Results: For all the three pairwise comparisons among CRC, polyps, and healthy controls, the highest classification accuracy of the NMR+MS data was obviously better than that from NMR or MS alone. For example, in the case of CRC Vs healthy controls, the NMR+MS data provided the highest classification accuracy of 0.95±0.05 after BVE, compared to 0.84±0.07 from NMR and 0.93±0.05 from MS. As expected, an excessive number of variables deteriorated statistical models, and there was an optimal number/range of variables which could produce the best statistical performance. Our approach was able to select an optimal set of metabolites for each pairwise comparison, and we achieved a comprehensive profile of altered metabolites from the combined (NMR+MS) data that were related to CRC and polyps, including those in glycolysis, the TCA cycle, amino acid metabolism, etc. Both NMR and MS detected metabolites contributed to the mixed panel of biomarker candidates, and thus both analytical platforms are very valuable methods to identify metabolic changes occurring in patients with CRC and polyps. The combined set of metabolite biomarkers should be helpful to comprehensively understand metabolite alterations of CRC and the mechanisms during disease progression. The majority of important NMR and MS metabolites were different from each other, evidencing that both NMR and MS can provide unique contributions to statistical modeling in metabolomics. Interestingly, CRC had the lowest adenosine levels while the controls had the highest levels. However, orotate (and some other metabolites) did not continuously increase or decrease from controls to polyps, and then to CRC, which indicates that disease progression could be a very complex process metabolically. Conclusions: The metabolic profiles of blood serum from CRC patients, polyps, and healthy controls were measured by NMR and LC-MS/MS and showed significant changes in metabolism with the onset of CRC. BVE-PLSDA identified optimal sets of metabolites that could be further validated for the diagnosis of CRC and polyps. The combination of NMR and MS showed significantly better statistical performance than NMR or MS alone. Therefore, we recommend the combined approach of NMR and MS through appropriate variable selection methods in metabolomics, especially for the purpose of discovering biomarker candidates. Notably, our approach is relatively universal and can be expanded to combine other analytical technologies. Citation Format: Lingli Deng, Haiwei Gu, Jiangjiang Zhu, Nagana Gowda, Danijel Djukovic, Daniel Raftery. Detecting colorectal cancer and polyps using nuclear magnetic resonance spectroscopy and mass spectrometry based metabolomics. [abstract]. In: Proceedings of the AACR Special Conference: Metabolism and Cancer; Jun 7-10, 2015; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Res 2016;14(1_Suppl):Abstract nr B50.

A clustering-based preprocessing method for the elimination of unwanted residuals in metabolomic data

IntroductionThe metabolome of a biological system is affected by multiple factors including factor of interest (e.g. metabolic perturbation due to disease) and unwanted factors or factors which are not primarily the focus of the study (e.g. batch effect, gender, and level of physical activity). Removal of these unwanted data variations is advantageous, as the unwanted variations may complicate biological interpretation of the data.ObjectivesWe aim to develop a new unwanted variations elimination (UVE) method called clustering-based unwanted residuals elimination (CURE) to reduce metabolic variation caused by unwanted/hidden factors in metabolomic data.MethodsA mean-centered metabolomic dataset can be viewed as a combination of a studied factor matrix and a residual matrix. The CURE method assumes that the residual should be normally distributed if it only contains inter-individual variation. However, if the residual forms multiple clusters in feature subspace of principal components analysis or partial least squares discriminant analysis, the residual may contain variation due to unwanted factors. This unwanted variation is removed by doing K-means data clustering and removal of means for each cluster from the residuals. The process is iterated until the residual no longer forms multiple clusters in feature subspace.ResultsThree simulated datasets and a human metabolomic dataset were used to demonstrate the performance of the proposed CURE method. CURE was found able to remove most of the variations caused by unwanted factors, while preserving inter-individual variation between samples.ConclusionThe CURE method can effectively remove unwanted data variation, and can serve as an alternative UVE method for metabolomic data.

Abstract B52: Targeted LC-MS/MS metabolic profiling for colon cancer progression monitoring

Introduction: Colorectal cancer (CRC) is one of the most prevalent cancers worldwide, and a major cause of human morbidity and mortality. A number of current efforts are focused on earlier detection of colon cancer using a variety of technologies including genomics, proteomics and metabolomics. Research focused on CRC disease status surveillance using metabolomics or other approaches has not been reported; Close monitoring of disease progression (DP) in CRC can be critical for patients9 prognosis management and treatment decisions. In this study we investigate a targeted LC-MS/MS approach for serum metabolic profiling to monitor and predict patient disease progression, using a panel of significantly altered metabolites as potential biomarkers. Methods: 59 serum samples from 21 CRC patients were analyzed, including 23 samples from DP patients and 36 from other CRC disease status (e.g., stable disease and complete remission). Chromatographic separations were performed via an Agilent HPLC system installed with two hydrophilic interaction chromatography (HILIC) columns, and then targeted data acquisition was performed in multiple-reaction-monitoring (MRM) mode using an AB Sciex QTrap 5500 mass spectrometer. We monitored 106 and 58 MRM transitions in negative and positive mode, respectively. Univariate and multivariate statistical analyses (such as the Mann- Whitney U-test and PLS-DA) were applied for metabolite biomarker discovery and model development on a selected set of promising biomarker candidates. Monte Carlo cross validation (MCCV) was performed to evaluate model robustness. Results and conclusion: LC-MS/MS targeted analysis provided a robust system for metabolic profiling of CRC patient disease status monitoring using serum samples. Targeted screening of 164 metabolites, representing more than 20 different classes (such as amino acids, carboxylic acids, pyridines, and etc.) and from 25 important metabolic pathways (e.g., TCA cycle, amino acid metabolism, purine and pyrimidine metabolism, and glycolysis, and etc.) was performed using both positive and negative ionization modes. 131 metabolites could be reproducibly detected in the serum samples, with an average CV of 7.1% measured in pooled serum quality control samples. After univariate analysis, 36 metabolites from different classes, such as monosaccharides, amino acids, carboxylic acids and nucleosides, showed a significant statistical difference (p Citation Format: Jiangjiang Zhu, Danijel Djukovic, Lingli Deng, Lingli Deng, Haiwei Gu, Farhan Himmati, Mohammad Abu Zaid, E. Gabriela Chiorean, E. Gabriela Chiorean, Daniel Raftery, Daniel Raftery. Targeted LC-MS/MS metabolic profiling for colon cancer progression monitoring. [abstract]. In: Proceedings of the AACR Special Conference: Metabolism and Cancer; Jun 7-10, 2015; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Res 2016;14(1_Suppl):Abstract nr B52.