Dietary patterns, untargeted metabolite profiles and their association with colorectal cancer risk

We investigated data-driven and hypothesis-driven dietary patterns and their association to plasma metabolite profiles and subsequent colorectal cancer (CRC) risk in 680 CRC cases and individually matched controls. Dietary patterns were identified from combined exploratory/confirmatory factor analysis. We assessed association to LC-MS metabolic profiles by random forest regression and to CRC risk by multivariable conditional logistic regression. Principal component analysis was used on metabolite features selected to reflect dietary exposures. Component scores were associated to CRC risk and dietary exposures using partial Spearman correlation. We identified 12 data-driven dietary patterns, of which a breakfast food pattern showed an inverse association with CRC risk (OR per standard deviation increase 0.89, 95% CI 0.80-1.00, p = 0.04). This pattern was also inversely associated with risk of distal colon cancer (0.75, 0.61-0.96, p = 0.01) and was more pronounced in women (0.69, 0.49-0.96, p = 0.03). Associations between meat, fast-food, fruit soup/rice patterns and CRC risk were modified by tumor location in women. Alcohol as well as fruit and vegetables associated with metabolite profiles (Q2 0.22 and 0.26, respectively). One metabolite reflecting alcohol intake associated with increased CRC risk, whereas three metabolites reflecting fiber, wholegrain, and fruit and vegetables associated with decreased CRC risk.

Untargeted plasma metabolomics and risk of colorectal cancer-an analysis nested within a large-scale prospective cohort

BACKGROUND

Colorectal cancer (CRC) is a leading cause of cancer-related death worldwide, but if discovered at an early stage, the survival rate is high. The aim of this study was to identify novel markers predictive of future CRC risk using untargeted metabolomics.

METHODS

This study included prospectively collected plasma samples from 902 CRC cases and 902 matched cancer-free control participants from the population-based Northern Sweden Health and Disease Study (NSHDS), which were obtained up to 26 years prior to CRC diagnosis. Using reverse-phase liquid chromatography-mass spectrometry (LC-MS), data comprising 5015 metabolic features were obtained. Conditional logistic regression was applied to identify potentially important metabolic features associated with CRC risk. In addition, we investigated if previously reported metabolite biomarkers of CRC risk could be validated in this study population.

RESULTS

In the univariable analysis, seven metabolic features were associated with CRC risk (using a false discovery rate cutoff of 0.25). Two of these could be annotated, one as pyroglutamic acid (odds ratio per one standard deviation increase = 0.79, 95% confidence interval, 0.70-0.89) and another as hydroxytigecycline (odds ratio per one standard deviation increase = 0.77, 95% confidence interval, 0.67-0.89). Associations with CRC risk were also found for six previously reported metabolic biomarkers of prevalent and/or incident CRC: sebacic acid (inverse association) and L-tryptophan, 3-hydroxybutyric acid, 9,12,13-TriHOME, valine, and 13-OxoODE (positive associations).

CONCLUSIONS

These findings suggest that although the circulating metabolome may provide new etiological insights into the underlying causes of CRC development, its potential application for the identification of individuals at higher risk of developing CRC is limited.

Circulating amino acid levels and colorectal cancer risk in the European Prospective Investigation into Cancer and Nutrition and UK Biobank cohorts

BACKGROUND

Amino acid metabolism is dysregulated in colorectal cancer patients; however, it is not clear whether pre-diagnostic levels of amino acids are associated with subsequent risk of colorectal cancer. We investigated circulating levels of amino acids in relation to colorectal cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) and UK Biobank cohorts.

METHODS

Concentrations of 13-21 amino acids were determined in baseline fasting plasma or serum samples in 654 incident colorectal cancer cases and 654 matched controls in EPIC. Amino acids associated with colorectal cancer risk following adjustment for the false discovery rate (FDR) were then tested for associations in the UK Biobank, for which measurements of 9 amino acids were available in 111,323 participants, of which 1221 were incident colorectal cancer cases.

RESULTS

Histidine levels were inversely associated with colorectal cancer risk in EPIC (odds ratio [OR] 0.80 per standard deviation [SD], 95% confidence interval [CI] 0.69-0.92, FDR P-value=0.03) and in UK Biobank (HR 0.93 per SD, 95% CI 0.87-0.99, P-value=0.03). Glutamine levels were borderline inversely associated with colorectal cancer risk in EPIC (OR 0.85 per SD, 95% CI 0.75-0.97, FDR P-value=0.08) and similarly in UK Biobank (HR 0.95, 95% CI 0.89-1.01, P=0.09) In both cohorts, associations changed only minimally when cases diagnosed within 2 or 5 years of follow-up were excluded.

CONCLUSIONS

Higher circulating levels of histidine were associated with a lower risk of colorectal cancer in two large prospective cohorts. Further research to ascertain the role of histidine metabolism and potentially that of glutamine in colorectal cancer development is warranted.

Pan-cancer analysis of pre-diagnostic blood metabolite concentrations in the European Prospective Investigation into Cancer and Nutrition

BACKGROUND

Epidemiological studies of associations between metabolites and cancer risk have typically focused on specific cancer types separately. Here, we designed a multivariate pan-cancer analysis to identify metabolites potentially associated with multiple cancer types, while also allowing the investigation of cancer type-specific associations.

METHODS

We analysed targeted metabolomics data available for 5828 matched case-control pairs from cancer-specific case-control studies on breast, colorectal, endometrial, gallbladder, kidney, localized and advanced prostate cancer, and hepatocellular carcinoma nested within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. From pre-diagnostic blood levels of an initial set of 117 metabolites, 33 cluster representatives of strongly correlated metabolites and 17 single metabolites were derived by hierarchical clustering. The mutually adjusted associations of the resulting 50 metabolites with cancer risk were examined in penalized conditional logistic regression models adjusted for body mass index, using the data-shared lasso penalty.

RESULTS

Out of the 50 studied metabolites, (i) six were inversely associated with the risk of most cancer types: glutamine, butyrylcarnitine, lysophosphatidylcholine a C18:2, and three clusters of phosphatidylcholines (PCs); (ii) three were positively associated with most cancer types: proline, decanoylcarnitine, and one cluster of PCs; and (iii) 10 were specifically associated with particular cancer types, including histidine that was inversely associated with colorectal cancer risk and one cluster of sphingomyelins that was inversely associated with risk of hepatocellular carcinoma and positively with endometrial cancer risk.

CONCLUSIONS

These results could provide novel insights for the identification of pathways for cancer development, in particular those shared across different cancer types.

Lifestyle correlates of eight breast cancer-related metabolites: a cross-sectional study within the EPIC cohort

BACKGROUND

Metabolomics is a promising molecular tool for identifying novel etiological pathways leading to cancer. In an earlier prospective study among pre- and postmenopausal women not using exogenous hormones, we observed a higher risk of breast cancer associated with higher blood concentrations of one metabolite (acetylcarnitine) and a lower risk associated with higher blood concentrations of seven others (arginine, asparagine, phosphatidylcholines (PCs) aa C36:3, ae C34:2, ae C36:2, ae C36:3, and ae C38:2).

METHODS

To identify determinants of these breast cancer-related metabolites, we conducted a cross-sectional analysis to identify their lifestyle and anthropometric correlates in 2358 women, who were previously included as controls in case-control studies nested within the European Prospective Investigation into Cancer and Nutrition cohort and not using exogenous hormones at blood collection. Associations of each metabolite concentration with 42 variables were assessed using linear regression models in a discovery set of 1572 participants. Significant associations were evaluated in a validation set (n = 786).

RESULTS

For the metabolites previously associated with a lower risk of breast cancer, concentrations of PCs ae C34:2, C36:2, C36:3, and C38:2 were negatively associated with adiposity and positively associated with total and saturated fat intakes. PC ae C36:2 was also negatively associated with alcohol consumption and positively associated with two scores reflecting adherence to a healthy lifestyle. Asparagine concentration was negatively associated with adiposity. Arginine and PC aa C36:3 concentrations were not associated to any of the factors examined. For the metabolite previously associated with a higher risk of breast cancer, acetylcarnitine, a positive association with age was observed.

CONCLUSIONS

These associations may indicate possible mechanisms underlying associations between lifestyle and anthropometric factors, and risk of breast cancer. Further research is needed to identify potential non-lifestyle correlates of the metabolites investigated.

Density of CD3+ and CD8+ Cells in the Microenvironment of Colorectal Cancer according to Prediagnostic Physical Activity

BACKGROUND

Physical activity is associated not only with a decreased risk of developing colorectal cancer but also with improved survival. One putative mechanism is the infiltration of immune cells in the tumor microenvironment. Experimental findings suggest that physical activity may mobilize immune cells to the tumor. We hypothesized that higher levels of physical activity prior to colorectal cancer diagnosis are associated with higher densities of tumor-infiltrating T-lymphocytes in colorectal cancer patients.

METHODS

The study setting was a northern Swedish population-based cohort, including 109,792 participants with prospectively collected health- and lifestyle-related data. For 592 participants who later developed colorectal cancer, archival tumor tissue samples were used to assess the density of CD3⁺ and CD8⁺ cytotoxic T cells by IHC. Odds ratios for associations between self-reported, prediagnostic recreational physical activity and immune cell infiltration were estimated by ordinal logistic regression.

RESULTS

Recreational physical activity >3 times per week was associated with a higher density of CD8⁺ T cells in the tumor front and center compared with participants reporting no recreational physical activity. Odds ratios were 2.77 (95% CI, 1.21-6.35) and 2.85 (95% CI, 1.28-6.33) for the tumor front and center, respectively, after adjustment for sex, age at diagnosis, and tumor stage. The risk estimates were consistent after additional adjustment for several potential confounders. For CD3, no clear associations were found.

CONCLUSIONS

Physical activity may promote the infiltration of CD8⁺ immune cells in the tumor microenvironment of colorectal cancer.

IMPACT

The study provides some evidence on how physical activity may alter the prognosis in colorectal cancer.

A New Pipeline for the Normalization and Pooling of Metabolomics Data

Pooling metabolomics data across studies is often desirable to increase the statistical power of the analysis. However, this can raise methodological challenges as several preanalytical and analytical factors could introduce differences in measured concentrations and variability between datasets. Specifically, different studies may use variable sample types (e.g., serum versus plasma) collected, treated, and stored according to different protocols, and assayed in different laboratories using different instruments. To address these issues, a new pipeline was developed to normalize and pool metabolomics data through a set of sequential steps: (i) exclusions of the least informative observations and metabolites and removal of outliers; imputation of missing data; (ii) identification of the main sources of variability through principal component partial R-square (PC-PR2) analysis; (iii) application of linear mixed models to remove unwanted variability, including samples' originating study and batch, and preserve biological variations while accounting for potential differences in the residual variances across studies. This pipeline was applied to targeted metabolomics data acquired using Biocrates AbsoluteIDQ kits in eight case-control studies nested within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. Comprehensive examination of metabolomics measurements indicated that the pipeline improved the comparability of data across the studies. Our pipeline can be adapted to normalize other molecular data, including biomarkers as well as proteomics data, and could be used for pooling molecular datasets, for example in international consortia, to limit biases introduced by inter-study variability. This versatility of the pipeline makes our work of potential interest to molecular epidemiologists.

Comparison of Methods for Feature Selection in Clustering of High-Dimensional RNA-Sequencing Data to Identify Cancer Subtypes

Cancer subtype identification is important to facilitate cancer diagnosis and select effective treatments. Clustering of cancer patients based on high-dimensional RNA-sequencing data can be used to detect novel subtypes, but only a subset of the features (e.g., genes) contains information related to the cancer subtype. Therefore, it is reasonable to assume that the clustering should be based on a set of carefully selected features rather than all features. Several feature selection methods have been proposed, but how and when to use these methods are still poorly understood. Thirteen feature selection methods were evaluated on four human cancer data sets, all with known subtypes (gold standards), which were only used for evaluation. The methods were characterized by considering mean expression and standard deviation (SD) of the selected genes, the overlap with other methods and their clustering performance, obtained comparing the clustering result with the gold standard using the adjusted Rand index (ARI). The results were compared to a supervised approach as a positive control and two negative controls in which either a random selection of genes or all genes were included. For all data sets, the best feature selection approach outperformed the negative control and for two data sets the gain was substantial with ARI increasing from (-0.01, 0.39) to (0.66, 0.72), respectively. No feature selection method completely outperformed the others but using the dip-rest statistic to select 1000 genes was overall a good choice. The commonly used approach, where genes with the highest SDs are selected, did not perform well in our study.

Combining epigenetic and clinicopathological variables improves specificity in prognostic prediction in clear cell renal cell carcinoma

BACKGROUND

Metastasized clear cell renal cell carcinoma (ccRCC) is associated with a poor prognosis. Almost one-third of patients with non-metastatic tumors at diagnosis will later progress with metastatic disease. These patients need to be identified already at diagnosis, to undertake closer follow up and/or adjuvant treatment. Today, clinicopathological variables are used to risk classify patients, but molecular biomarkers are needed to improve risk classification to identify the high-risk patients which will benefit most from modern adjuvant therapies. Interestingly, DNA methylation profiling has emerged as a promising prognostic biomarker in ccRCC. This study aimed to derive a model for prediction of tumor progression after nephrectomy in non-metastatic ccRCC by combining DNA methylation profiling with clinicopathological variables.

METHODS

A novel cluster analysis approach (Directed Cluster Analysis) was used to identify molecular biomarkers from genome-wide methylation array data. These novel DNA methylation biomarkers, together with previously identified CpG-site biomarkers and clinicopathological variables, were used to derive predictive classifiers for tumor progression.

RESULTS

The "triple classifier" which included both novel and previously identified DNA methylation biomarkers together with clinicopathological variables predicted tumor progression more accurately than the currently used Mayo scoring system, by increasing the specificity from 50% in Mayo to 64% in our triple classifier at 85% fixed sensitivity. The cumulative incidence of progress (_pCIP_5yr) was 7.5% in low-risk vs 44.7% in high-risk in M0 patients classified by the triple classifier at diagnosis.

CONCLUSIONS

The triple classifier panel that combines clinicopathological variables with genome-wide methylation data has the potential to improve specificity in prognosis prediction for patients with non-metastatic ccRCC.

Abstract 5475: Evaluation of feature selection methods used for cluster analysis in identification of novel cancer subtypes

Background: RNA-seq data from tumor samples can be used to identify novel cancer subtypes using cluster analysis. The number of features is often large compared to the number of samples and different clusters can appear in different subsets of the feature space. Feature selection techniques are therefore commonly used to reduce the dimension and remove redundant and irrelevant features before performing cluster analysis. An abundance of feature selection methods have been proposed in the literature, but it is unclear how the ability to identify novel subtypes of cancer using cluster analysis is affected by choice of feature selection method.

Method: We evaluated 13 feature selection methods on four publicly available cancer data sets from The Cancer Genome Atlas. RNA-seq data and associated clinical data were retrieved from Broad institute GDAC Firehose. Overlap and characterization of the highest ranked features (i.e. genes) were studied for top 100, 1000 and 3000 genes. Performance was measured by comparing known cancer subtypes to partitions obtained using hierarchical clustering with Euclidean distance and Ward´s linkage, based on the selected features. The result was compared to both a random selection (negative control) and a supervised approach (positive control) using adjusted Rand index.

Results: The relative performance of the feature selection methods varied heavily depending on number of included genes and the data set used. Based on all data, gene selection using the Dip-test statistic and the Bimodality index generated the overall highest clustering performance. The overlap of selected genes between the Bimodality index, the Dip-test and the supervised approach was relatively low. The Dip-test and Bimodality index tended to favor genes with relatively low expression values.

Conclusions: The choice of feature selection method can have a huge impact on the ability to identify cancer subtypes using cluster analysis, and the relative performance is highly dependent on the data. Low overlap of selected genes between the highest ranked methods suggests that the methods identify genes that contain complementary information, and that it might be beneficial to combine two or more feature selection methods.

Citation Format: Linda Vidman, David Källberg, Patrik Rydén. Evaluation of feature selection methods used for cluster analysis in identification of novel cancer subtypes [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5475.

Cluster analysis on high dimensional RNA-seq data with applications to cancer research - An evaluation study

BACKGROUND

Clustering of gene expression data is widely used to identify novel subtypes of cancer. Plenty of clustering approaches have been proposed, but there is a lack of knowledge regarding their relative merits and how data characteristics influence the performance. We evaluate how cluster analysis choices affect the performance by studying four publicly available human cancer data sets: breast, brain, kidney and stomach cancer. In particular, we focus on how the sample size, distribution of subtypes and sample heterogeneity affect the performance.

RESULTS

In general, increasing the sample size had limited effect on the clustering performance, e.g. for the breast cancer data similar performance was obtained for n = 40 as for n = 330. The relative distribution of the subtypes had a noticeable effect on the ability to identify the disease subtypes and data with disproportionate cluster sizes turned out to be difficult to cluster. Both the choice of clustering method and selection method affected the ability to identify the subtypes, but the relative performance varied between data sets, making it difficult to rank the approaches. For some data sets, the performance was substantially higher when the clustering was based on data from only one sex compared to data from a mixed population. This suggests that homogeneous data are easier to cluster than heterogeneous data and that clustering males and females individually may be beneficial and increase the chance to detect novel subtypes. It was also observed that the performance often differed substantially between females and males.

CONCLUSIONS

The number of samples seems to have a limited effect on the performance while the heterogeneity, at least with respect to sex, is important for the performance. Hence, by analyzing the genders separately, the possible loss caused by having fewer samples could be outweighed by the benefit of a more homogeneous data.

Gene expression profiles define molecular subtypes of prostate cancer bone metastases with different outcomes and morphology traceable back to the primary tumor

Bone metastasis is the lethal end-stage of prostate cancer (PC), but the biology of bone metastases is poorly understood. The overall aim of this study was therefore to explore molecular variability in PC bone metastases of potential importance for therapy. Specifically, genome-wide expression profiles of bone metastases from untreated patients (n = 12) and patients treated with androgen-deprivation therapy (ADT, n = 60) were analyzed in relation to patient outcome and to morphological characteristics in metastases and paired primary tumors. Principal component analysis and unsupervised classification were used to identify sample clusters based on mRNA profiles. Clusters were characterized by gene set enrichment analysis and related to histological and clinical parameters using univariate and multivariate statistics. Selected proteins were analyzed by immunohistochemistry in metastases and matched primary tumors (n = 52) and in transurethral resected prostate (TUR-P) tissue of a separate cohort (n = 59). Three molecular subtypes of bone metastases (MetA-C) characterized by differences in gene expression pattern, morphology, and clinical behavior were identified. MetA (71% of the cases) showed increased expression of androgen receptor-regulated genes, including prostate-specific antigen (PSA), and glandular structures indicating a luminal cell phenotype. MetB (17%) showed expression profiles related to cell cycle activity and DNA damage, and a pronounced cellular atypia. MetC (12%) exhibited enriched stroma-epithelial cell interactions. MetB patients had the lowest serum PSA levels and the poorest prognosis after ADT. Combined analysis of PSA and Ki67 immunoreactivity (proliferation) in bone metastases, paired primary tumors, and TUR-P samples was able to differentiate MetA-like (high PSA, low Ki67) from MetB-like (low PSA, high Ki67) tumors and demonstrate their different prognosis. In conclusion, bone metastases from PC patients are separated based on gene expression profiles into molecular subtypes with different morphology, biology, and clinical outcome. These findings deserve further exploration with the purpose of improving treatment of metastatic PC.

Epidemiology and ecology of tularemia in Sweden, 1984-2012

Geographic distribution of cases was correlated with the locations of lakes and rivers.

The zoonotic disease tularemia is endemic in large areas of the Northern Hemisphere, but research is lacking on patterns of spatial distribution and connections with ecologic factors. To describe the spatial epidemiology of and identify ecologic risk factors for tularemia incidence in Sweden, we analyzed surveillance data collected over 29 years (1984–2012). A total of 4,830 cases were notified, of which 3,524 met all study inclusion criteria. From the first to the second half of the study period, mean incidence increased 10-fold, from 0.26/100,000 persons during 1984–1998 to 2.47/100,000 persons during 1999–2012 (p<0.001). The incidence of tularemia was higher than expected in the boreal and alpine ecologic regions (p<0.001), and incidence was positively correlated with the presence of lakes and rivers (p<0.001). These results provide a comprehensive epidemiologic description of tularemia in Sweden and illustrate that incidence is higher in locations near lakes and rivers.