An Extensive MicroRNA-Mediated Network of RNA-RNA Interactions Regulates Established Oncogenic Pathways in Glioblastoma
MicroRNAs (miRs) are small non-coding RNAs of 21-25 nucleotides that bind to partially complementary sites in their target RNAs, directly inducing RNA degradation and mRNA translational repression. A growing body of evidence has linked miRs to tumorigenesis and tumor progression, suggesting their potential value as biomarkers and as targets for therapeutic intervention. Currently, miRs are known to regulate tumor cell growth, and their expression profiles are used to classify tumors and to differentiate between molecular tumor subtypes. Conversely, mRNAs have long been thought to be passive carriers of genetic information and their regulatory roles as RNAs in normal biological process and development of disease such as cancer were largely disregarded by scientists.
We have recently uncovered a new post-transcriptional regulation layer called the miR-Program mediated Regulatory (mPR) network. Studying regulation in glioblastoma, we showed that mRNAs can regulate, and be regulated by, other mRNAs by competing for limited pools of low expressed miRs. The rationale behind the mPR network is that when two mRNAs share a common set of miR regulators, increases in the number of transcripts of one mRNA will recruit (or sponge up) more of the available miRs and induce corresponding increases in the number of translatable transcripts of the other mRNA, and vice versa.
Building an mPR network relies on high-precision prediction of miR targets, but current methods are notoriously unreliable. A computational method Cupid is specifically designed to produce very low false-positive rates, possibly at the expense of false-negative rates, by using stringent, integrative selection criteria. For genome-wide computational screening of mPR interactions, we developed Hermes to integrate large-scale gene and miR expression data with Cupid-inferred common miR programs. Specifically, Hermes relies on the mutual information estimated via adaptive portioning approach to measure the nonlinear relationship between two RNAs if they are predicted to share a significantly large common miR program.
In this study, bioinformatics predictions, followed by experimental validation in cell lines, offer evidence supporting the conclusion that mRNAs can play important roles in gene regulation and affect disease pathogenesis via mPR interactions by targeting tumor-suppressors or oncogenes. Joint analysis of multiple tumor types suggests that while almost all mPR interactions are implemented by miR programs with tumor type-specific expression, hundreds of thousands of interactions that regulate gene expression through these context-specific implementations are themselves context independent.
Title:Mutational landscape of esophageal cancer
Abstract:With its low 5-year survival rate and rising incidence, esophageal cancer has become the focus of several recent studies using next-generation technologies. In order to profile the mutational landscape of esophageal cancer and identify the driver mutations, we sequenced the exomes of 85 cancer subjects. All tumor samples were obtained at primary diagnosis and resected during surgeries, with the normal tissues from adjacent areas during the same surgery. Whole exome sequencing was performed on the samples, with an average depth of coverage ~ 100X. After applying standard pipelines in sequencing analysis (BWA and GATK) on these tumor/normal pairs, we will call somatic mutations with MuTect, which is able to detect mutations carried in a minority of clones. We will also infer copy number alternations (CNAs) using XHMM. We will then calculate driver gene scores based on recurrency and background mutation rate. Finally, we will correlate mutation profiles with post-surgery survival time and explore the association between the two