Context-specific gene regulatory networks subdivide intrinsic subtypes of breast cancer

Identification of specific microRNA-messenger RNA regulation pairs in four subtypes of breast cancer

Four subtypes of breast cancer, luminal A, luminal B, basal-like, human epidermal growth factor receptor-enriched, have been identified based on gene expression profiles of human tumours. The goal of this study is to find whether the same groups' genes would exhibit different networks among the four subtypes. Differential expressed genes between each of the four subtypes and the normal samples were identified. The overlaps between the four groups of differentially expressed genes were used to construct regulations networks for each of the four subtypes. Univariate and multivariate Cox regressions were employed to test the genes in the four regulation networks. This study demonstrated that the common genes in four subtypes showed different regulation. Also, the hsa-miR-182 and decorin pair performs different functions among the four subtypes of breast cancer. The result indicated that heterogeneity of breast cancer is not only reflected in the different expression patterns among different genes, but also in the different regulatory networks of the same group of genes.

A closer look at cross-validation for assessing the accuracy of gene regulatory networks and models

Cross-validation (CV) is a technique to assess the generalizability of a model to unseen data. This technique relies on assumptions that may not be satisfied when studying genomics datasets. For example, random CV (RCV) assumes that a randomly selected set of samples, the test set, well represents unseen data. This assumption doesn’t hold true where samples are obtained from different experimental conditions, and the goal is to learn regulatory relationships among the genes that generalize beyond the observed conditions. In this study, we investigated how the CV procedure affects the assessment of supervised learning methods used to learn gene regulatory networks (or in other applications). We compared the performance of a regression-based method for gene expression prediction estimated using RCV with that estimated using a clustering-based CV (CCV) procedure. Our analysis illustrates that RCV can produce over-optimistic estimates of the model’s generalizability compared to CCV. Next, we defined the ‘distinctness’ of test set from training set and showed that this measure is predictive of performance of the regression method. Finally, we introduced a simulated annealing method to construct partitions with gradually increasing distinctness and showed that performance of different gene expression prediction methods can be better evaluated using this method.

Identifying gene set association enrichment using the coefficient of intrinsic dependence

Gene set testing problem has become the focus of microarray data analysis. A gene set is a group of genes that are defined by a priori biological knowledge. Several statistical methods have been proposed to determine whether functional gene sets express differentially (enrichment and/or deletion) in variations of phenotypes. However, little attention has been given to analyzing the dependence structure among gene sets. In this study, we have proposed a novel statistical method of gene set association analysis to identify significantly associated gene sets using the coefficient of intrinsic dependence. The simulation studies show that the proposed method outperforms the conventional methods to detect general forms of association in terms of control of type I error and power. The correlation of intrinsic dependence has been applied to a breast cancer microarray dataset to quantify the un-supervised relationship between two sets of genes in the tumor and non-tumor samples. It was observed that the existence of gene-set association differed across various clinical cohorts. In addition, a supervised learning was employed to illustrate how gene sets, in signaling transduction pathways or subnetworks regulated by a set of transcription factors, can be discovered using microarray data. In conclusion, the coefficient of intrinsic dependence provides a powerful tool for detecting general types of association. Hence, it can be useful to associate gene sets using microarray expression data. Through connecting relevant gene sets, our approach has the potential to reveal underlying associations by drawing a statistically relevant network in a given population, and it can also be used to complement the conventional gene set analysis.