Correlation-Based Network Generation, Visualization, and Analysis as a Powerful Tool in Biological Studies: A Case Study in Cancer Cell Metabolism

Abstract

In the last decade vast data sets are being generated in biological and medical studies. The challenge lies in their summary, complexity reduction, and interpretation. Correlation-based networks and graph-theory based properties of this type of networks can be successfully used during this process. However, the procedure has its pitfalls and requires specific knowledge that often lays beyond classical biology and includes many computational tools and software. Here we introduce one of a series of methods for correlation-based network generation and analysis using freely available software. The pipeline allows the user to control each step of the network generation and provides flexibility in selection of correlation methods and thresholds. The pipeline was implemented on published metabolomics data of a population of human breast carcinoma cell lines MDA-MB-231 under two conditions: normal and hypoxia. The analysis revealed significant differences between the metabolic networks in response to the tested conditions. The network under hypoxia had 1.7 times more significant correlations between metabolites, compared to normal conditions. Unique metabolic interactions were identified which could lead to the identification of improved markers or aid in elucidating the mechanism of regulation between distantly related metabolites induced by the cancer growth.

Figure 1

Correlation-based network, pipeline flowchart.

Figure 2

Correlation-based networks of metabolite data sets. (a) Original network under control conditions, (b) original network under hypoxic conditions, (c) network of unique relationships under control conditions compared to hypoxic conditions, and (d) network of unique relationships under hypoxic conditions compared to control conditions. Metabolic profiling of breast cancer cells under control and hypoxia (30 samples each) was used for pairwise correlation analysis between metabolites and network-view production. The data used to generate the network is from Kotze et al., 2013 [20]. Each vertex represents a metabolite; each edge represents a significant correlation between pairs of metabolites across samples. Vertex colors reflect biochemical classes: amino acids and N-compounds (blue), sugars and sugar alcohols (orange), and carboxylic acids (green). Vertex size reflects degree.