Lipidomic profiles and colorectal cancer risk
Using the stored plasma samples, we will measure up to ~1,000 molecular species of lipids at the Lipidomic/Metabolomic Center at Veterans Affairs Medical Center (VAMC) in Northport, NY, a VAMC affiliated with Stony Brook University (M. Del Poeta, PI). The laboratory is equipped with Agilent 6545 and Agilent 6550 QTOF LC-MS high resolution spectrometers, each equipped with HPLC. These instruments have dual ESI sources, and they allow detection of all ionic lipids, and most neutral, but ionizable ones. The instruments have been used in previous studies based on multiple reaction monitoring and other targeted lipidomics methods. The proposed panel will provide quantification of lipid species across 15 lipid classes, including cholesteryl esters, monoacylglycerols, diacylglycerols, triglycerides, free fatty acids, phosphatidylcholines, phosphatidylethanolamines, phosphatidylinositols, lysophosphatidylcholines, lysophosphatidylethanolamines, ceramides, dihydroceramides, hexosylceramides, lactosylceramides, and sphingomyelins.
Untargeted lipidomic data will be transformed, normalized, and analyzed using MetaboAnalyst 4.0. The Benjamini–Hochberg procedure will be used to control the false discovery rate (FDR), and the markers with FDR ≤ 0.1 and absolute log2 fold change ≥ 1.5 will be considered as significant and biologically relevant. Individual metabolites will be compared between groups with non-parametric Mann-Whitney Wilcoxon tests at an α-level of 0.05, and receiver operating characteristic (ROC) curves, area under the curves (AUC), and confidence intervals will be generated. Clinical indices will be used as input variables to build a predictive model (i.e., decision tree) by recursive partitioning, using the Classification and Regression Trees (CART) algorithm50 implemented in the R package RPART. The tree model will identify a set of predictive features (branch conditions) that best classified subjects as cases and controls. The tree split points will be determined by the Gini index with the minimum leaf size = 10. A tenfold cross-validation method will be used to tune the tree model and evaluate its prediction accuracy. To avoid overfitting, the tree will be pruned back to the smallest size while minimizing the cross-validated error. The classification accuracy of the tree to determine group membership (e.g., cases vs. controls) will be assessed using the area under the ROC curve. The importance of each clinical or epidemiological variable will be assessed by the decrease of prediction accuracy when such a feature is omitted in the model, based on two measurement metrics: Gini importance or Mean Decrease Impurity, and permutation importance or Mean Decrease Accuracy.
Spearman's rank correlation coefficients will be calculated to assess the correlations between each tree model among controls. For the nested case-control analysis of both individual markers and lipidomic profiles, we will use conditional logistic regression to calculate odds ratios and 95% confidence intervals. Stratified analyses will be conducted by site, gender, age, and other characteristics.
Colorectal cancer is one of the most common cancers and second most common cause of cancer death globally in 2022. According to the WCRF Continuous Update Project Expert Report 2018 “Diet, nutrition, physical activity and colorectal cancer”, there is strong evidence that colorectal cancer is linked to the Western diet, such as consumption of red and processed meat, alcoholic drinks, and being overweight or obese. However, the underlying mechanisms are not fully understood.
Recent progress in analytical technologies such as mass spectrometry (MS) has allowed to investigate the human metabolome in large-scale epidemiological studies. Lipids play key roles in biological processes such as energy metabolism, cell membrane structure, cell signalling and proliferation, and the dysregulation of lipid metabolism is often one of the first stages of cancer development. A potential relationship between circulating lipid metabolites and colorectal cancer has also been supported by epidemiological studies of several individual lipids to colorectal cancer risk factors such as diet, cholesterol levels, adiposity, alcohol consumption, smoking, diabetes, and other lifestyle measures.
To date, few prospective studies have been conducted to investigate the association between lipid metabolites. Recent studies on colorectal cancer were based on limited panels of lipids, and have provided inconsistent results. Therefore, there is a need for further larger-scale studies including a comprehensive panel of lipid markers to elucidate the role of lipid metabolites in colorectal cancer development.
The project will consist of three Specific Aims.
Specific Aim 1: To measure a large spectrum of lipids in plasma samples of cases of colorectal cancer and controls from PLCO, and to identify lipidomic profiles in controls. This aim will be achieved in collaboration with a specialized laboratory affiliated with Stony Brook University.
Specific Aim 2: To investigate the association between lipidomic profiles identified in Aim 1 and colorectal cancer risk. We will conduct a nested case-control analysis within the PLCO and estimate the association between individual lipids and lipid profiles and CRC risk via multivariate regression analysis.
Specific Aim 3: To explore the mediation role of lipids and lipidomic profiles in the association between high-risk food items and food groups, such as red meat, processed meat, and dairy products, and CRC risk.
Paolo Boffetta (State University of New York at Stony Brook)
Maurizo Del Poeta (State University of New York at Stony Brook)