Evaluating Ultra-long-Chain Fatty Acids as Biomarkers of Colorectal Cancer Risk

Abstract

Background: Cross-sectional studies reported a novel set of hydroxylated ultra-long-chain fatty acids (ULCFA) that were present at significantly lower levels in colorectal cancer cases than controls. Follow-up studies suggested that these molecules were potential biomarkers of protective exposure for colorectal cancer. To test the hypothesis that ULCFAs reflect causal pathways, we measured their levels in prediagnostic serum from incident colorectal cancer cases and controls.Methods: Serum from 95 colorectal cancer patients and 95 matched controls was obtained from the Italian arm of the European Prospective Investigation into Cancer and Nutrition cohort and analyzed by liquid chromatography-high-resolution mass spectrometry. Levels of 8 ULCFAs were compared between cases and controls with paired t tests and a linear model that used time to diagnosis (TTD) to determine whether case-control differences were influenced by disease progression.

Data-adaptive pipeline for filtering and normalizing metabolomics data

“…One of these datasets represents the metabolomes of 4.7-mm punches from archived neonatal blood spots (NBS) of 400 control subjects that were obtained for the California Childhood Leukemia Study (Metayer et al, 2013) from the California birth registry , as described in Petrick et al (2017). The second dataset contains the metabolomes of 122 serum samples from incident colorectal cancer (CRC) casecontrol pairs as described in Perttula et al (2016). For comparison purposes, we define a binary biological factor of interest for each dataset, i.e., breastfeeding for the NBS (newborn infants are fed at least once prior to collection of NBS) and CRC diagnosis for the serum samples.…”

“…Over 20,000 features were detected in the 122 serum samples (61 incident cases, 61 controls) that were analyzed in two batches (Perttula et al, 2016). Fifty-two of the samples in the first batch were contaminated by additives that resulted from gelled serum, thereby introducing an additional known source of technical variation into the data.…”

“…Sample age was not considered as a source of variation, because all specimens were collected during a small time window and were stored in liquid nitrogen soon after collection. As positive controls for CRC, we used eight ultra-long-chain fatty acids (ULCFA 446,448,466,468,492,494,538,594), that had previously been shown to be present at lower levels in CRC cases than controls in several cross-sectional studies as well as in this set of samples (summarized by Perttula et al (2016)).…”

“…ULCFA 448 was retained by the traditional pipeline but not the data-adaptive pipeline because it had a small ICC value (0.18) and unsurprisingly ranked 1210th in terms of p-values. (Petrick et al, 2017) and serum (Perttula et al, 2016) datasets.…”

“…Traditional and data-adaptive filtering and normalization, NBS and serum datasets. Principal component plots for the NBS(Petrick et al, 2017) and serum(Perttula et al, 2016) datasets after traditional and data-adaptive filtering and normalization. Samples are color-coded by batch.…”