Dataset Information

Identifying stage-specific protein subnetworks for colorectal cancer.

ABSTRACT:

Background

In recent years, many algorithms have been developed for network-based analysis of differential gene expression in complex diseases. These algorithms use protein-protein interaction (PPI) networks as an integrative framework and identify subnetworks that are coordinately dysregulated in the phenotype of interest.

Motivation

While such dysregulated subnetworks have demonstrated significant improvement over individual gene markers for classifying phenotype, the current state-of-the-art in dysregulated subnetwork discovery is almost exclusively limited to binary phenotype classes. However, many clinical applications require identification of molecular markers for multiple classes.

Approach

We consider the problem of discovering groups of genes whose expression signatures can discriminate multiple phenotype classes. We consider two alternate formulations of this problem (i) an all-vs-all approach that aims to discover subnetworks distinguishing all classes, (ii) a one-vs-all approach that aims to discover subnetworks distinguishing each class from the rest of the classes. For the one-vs-all formulation, we develop a set-cover based algorithm, which aims to identify groups of genes such that at least one gene in the group exhibits differential expression in the target class.

Results

We test the proposed algorithms in the context of predicting stages of colorectal cancer. Our results show that the set-cover based algorithm identifying "stage-specific" subnetworks outperforms the all-vs-all approaches in classification. We also investigate the merits of utilizing PPI networks in the search for multiple markers, and show that, with correct parameter settings, network-guided search improves performance. Furthermore, we show that assessing statistical significance when selecting features greatly improves classification performance.

SUBMITTER: Erten S

PROVIDER: S-EPMC3504924 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:The progress in the discovery and validation of metabolite biomarkers for the detection of colorectal cancer (CRC) has been hampered by the lack of reproducibility between study cohorts. The majority of discovery-phase biomarker studies have used patient blood samples to identify disease-related metabolites, but this pre-validation phase is confounded by non-specific disease influences on the metabolome. We therefore propose that metabolite biomarker discovery would have greater success and higher reproducibility for CRC if the discovery phase was conducted in tumor tissues, to find metabolites that have higher specificity to the metabolic consequences of the disease, that are then validated in blood samples. This would thereby eliminate any non-tumor and/or body response effects to the disease. In this study, we performed comprehensive untargeted metabolomics analyses on normal (adjacent) colon and tumor tissues from CRC patients, revealing tumor tissue-specific biomarkers (n = 39/group). We identified 28 highly discriminatory tumor tissue metabolite biomarkers of CRC by orthogonal partial least-squares discriminant analysis (OPLS-DA) and univariate analyses (VIP > 1.5, p < 0.05). A stepwise selection procedure was used to identify nine metabolites that were the most predictive of CRC with areas under the curve (AUCs) of >0.96, using various models. We further identified five biomarkers that were specific to the anatomic location of tumors in the colon (n = 236). The combination of these five metabolites (S-adenosyl-L-homocysteine, formylmethionine, fucose 1-phosphate, lactate, and phenylalanine) demonstrated high differentiative capability for left- and right-sided colon cancers at stage I by internal cross-validation (AUC = 0.804, 95% confidence interval, CI 0.670-0.940). This study thus revealed nine discriminatory biomarkers of CRC that are now poised for external validation in a future independent cohort of samples. We also discovered a discrete metabolic signature to determine the anatomic location of the tumor at the earliest stage, thus potentially providing clinicians a means to identify individuals that could be triaged for additional screening regimens.