microRNAs are key regulators of the human transcriptome across a number of diverse biological processes, such as development, aging and cancer, where particular miRNAs have been identified as tumour suppressive and oncogenic. In this work, we elucidate, in a comprehensive manner, across 15 epithelial cancer types comprising 7316 clinical samples from the Cancer Genome Atlas, the association of miRNA expression and target regulation with the phenotypic hallmarks of cancer. Utilising penalised regression techniques to integrate transcriptomic, methylation and mutation data, we find evidence for a complex map of interactions underlying the relationship of miRNA regulation and the hallmarks of cancer. This highlighted high redundancy for the oncomiR-1 cluster of oncogenic miRNAs, in particular hsa-miR-17-5p. In addition, we reveal extensive miRNA regulation of tumour suppressor genes such as PTEN, FAT4 and CDK12, uncovering an alternative mechanism of repression in the absence of mutation, methylation or copy number changes.

With the increase in next generation sequencing generating large amounts of genomic data, gene expression signatures are becoming critically important tools to interpret these data, and are poised to make a large impact on diagnosis, management and prognosis for a number of diseases. Increasingly, it is becoming crucial to establish whether the expression patterns and statistical properties of a set of genes, or signature, are conserved across datasets. Conversely, it is increasingly necessary to compare independent datasets with respect to the expression of established signatures reflecting their clinical or biological characteristics. In this work, we introduce the first protocol, sigQC, which enables a streamlined, systematic approach for the evaluation of gene signatures across different, independent, datasets. To facilitate accessibility, we implemented the protocol in an R package (https://cran.r-project.org/web/packages/sigQC/) designed for users with modest computational skills. SigQC has been adopted by us and collaborators in several basic biology and biomarker studies. The emphasis is in showing the basic but critical quality control steps involved in the generation and application of a clinically and biologically useful, transportable gene signature, including evaluating its expression, variability and structure. It begins with evaluating signature genes' expression and variability, then evaluates statistical properties of the distribution of their expression, and then computes various signature scoring metrics, and gives empirical estimates for the significance of each of these metrics. We demonstrate the application of this protocol, showing how the outputs created from sigQC may be used for the evaluation of gene signatures on large-scale gene expression datasets. cost of potentially increased noise from these non-linear relationships. S-scoring is based on a linear combination of z-scores, and combines the approaches of standardising the dataset with the directionality and flexibility of a linear model 30 . Thus, like a linear model, this scoring system, while it may be more flexible for defining dataset specific scores, often does not translate easily to new datasets or technologies. These methods are not tested explicitly in the current version of sigQC, however the metrics provided by sigQC provide a broad statistical assessment of the genes in a given signature across datasets and technologies, information which can be used to design more context-specific scoring techniques. MaterialsEquipment Hardware:  Personal computer, capable of running R version 3.3.0 or higher Software:  R version ≥ 3.3.0, available to install from https://www.r-project.org/  Bioconductor compatible with R version; installation instructions available from https://www.bioconductor.org/install/  sigQC package, available to download from https://cran.rproject.org/web/packages/sigQC/index.html  The following R packages are required as dependencies: MASS, lattice, KernSmooth, cluster, nnet, class, gridGraphics, biclust,...

Tumor heterogeneity includes variable and fluctuating oxygen concentrations, which result in the accumulation of hypoxic regions in most solid tumors. Tumor hypoxia leads to increased therapy resistance and has been linked to genomic instability. Here, we tested the hypothesis that exposure to levels of hypoxia that cause replication stress could increase APOBEC activity and the accumulation of APOBEC-mediated mutations. APOBEC-dependent mutational signatures have been well-characterized, although the physiological conditions which underpin them have not been described. We demonstrate that fluctuating/cyclic hypoxic conditions which lead to replication catastrophe induce the expression and activity of APOBEC3B. In contrast, stable/chronic hypoxic conditions which induce replication stress in the absence of DNA damage are not sufficient to induce APOBEC3B. Most importantly, the number of APOBEC-mediated mutations in patient tumors correlated with a hypoxia signature. Together, our data support the conclusion that hypoxia-induced replication catastrophe drives genomic instability in tumors, specifically through increasing the activity of APOBEC3B.