petal: A novel co-expression network modeling system

With the introduction of microarray technology to measure gene expression in the late 1990's and the advent of current next-generation whole-genome and whole-transcriptome sequencing technology, fast and robust tools are needed to examine, identify, and study relations between genes and proteins at the systems level. Networks provide effective models to study complex systems, including complex biological systems, such as gene and protein interaction networks. We introduce a novel approach to generate gene co-expression network models based on experimental gene expression measures. This approach includes statistical, mathematical, and biological considerations not highlighted by existing co-expression analysis tools. First, most high-throughput expression data are not normally distributed, yet many available network approaches are based on parametric methods. Here appropriate metrics are provided to generate statistically sound network models. Secondly, most biological networks are known to have approximate scale-free and small-world structure, and the biological networks built here follow both these two properties. Thirdly, to generate these small-world, scale-free network models, user-selected input parameters are not required, thereby leading to reproducible results. Lastly, this approach is designed for high-throughput whole-systems data. This method is implemented in the programming language R. Its application to several whole-genome experimental datasets has generated novel meaningful results useful for further biological investigation.