Inference of epigenetic subnetworks by Bayesian regression with the incorporation of prior information.

Changes in gene expression have been thought to play a crucial role in various types of cancer. With the advance of high-throughput experimental techniques, many genome-wide studies are underway to analyze underlying mechanisms that may drive the changes in gene expression. It has been observed that the change could arise from altered DNA methylation. However, the knowledge about the degree to which epigenetic changes might cause differences in gene expression in cancer is currently lacking. By considering the change of gene expression as the response of altered DNA methylation, we introduce a novel analytical framework to identify epigenetic subnetworks in which the methylation status of a set of highly correlated genes is predictive of a set of gene expression. By detecting highly correlated modules as representatives of the regulatory scenario underling the gene expression and DNA methylation, the dependency between DNA methylation and gene expression is explored by a Bayesian regression model with the incorporation of g-prior followed by a strategy of an optimal predictor subset selection. The subsequent network analysis indicates that the detected epigenetic subnetworks are highly biologically relevant and contain many verified epigenetic causal mechanisms. Moreover, a survival analysis indicates that they might be effective prognostic factors associated with patient survival time.

Expression-based analyses indicate a central role for hypoxia in driving tumor plasticity through microenvironment remodeling and chromosomal instability.

Can transcriptomic alterations drive the evolution of tumors? We asked if changes in gene expression found in all patients arise earlier in tumor development and can be relevant to tumor progression. Our analyses of non-mutated genes from the non-amplified regions of the genome of 158 triple-negative breast cancer (TNBC) cases identified 219 exclusively expression-altered (EEA) genes that may play important role in TNBC. Phylogenetic analyses of these genes predict a "punctuated burst" of multiple gene upregulation events occurring at early stages of tumor development, followed by minimal subsequent changes later in tumor progression. Remarkably, this punctuated burst of expressional changes is instigated by hypoxia-related molecular events, predominantly in two groups of genes that control chromosomal instability (CIN) and those that remodel tumor microenvironment (TME). We conclude that alterations in the transcriptome are not stochastic and that early-stage hypoxia induces CIN and TME remodeling to permit further tumor evolution.

A novel feature extraction scheme for prediction of protein-protein interaction sites.

Identifying protein-protein interaction (PPI) sites plays an important and challenging role in some topics of biology. Although many methods have been proposed, this problem is still far away to be solved. Here, a feature selection approach with an 11-sliding window and random forest algorithm is proposed, which is called DX-RF. This method has achieved an accuracy of 88.79%, recall of 82.09%, and precision of 85.76% with top-ranked 34 features on the Hetero test dataset and has 91.6% accuracy, 89.2% precision, 83.54% recall with top-ranked 25 features set on the Homo test dataset. Compared to other methods, the results indicate that the DX-RF method has a strong ability to select relevance features to get a higher performance. Moreover, in order to further understand protein interactions, feature analysis in this study is also performed.