A polygenic two-hit hypothesis for prostate cancer.

Prostate cancer is one of the most heritable cancers. Hundreds of germline polymorphisms have been linked to prostate cancer diagnosis and prognosis. Polygenic risk scores can predict genetic risk of a prostate cancer diagnosis. Although these scores inform the probability of developing a tumor, it remains unknown how germline risk influences the tumor molecular evolution. We cultivated a cohort of 1250 localized European-descent patients with germline and somatic DNA profiling. Men of European descent with higher genetic risk were diagnosed earlier and had less genomic instability and fewer driver genes mutated. Higher genetic risk was associated with better outcome. These data imply a polygenic "two-hit" model where germline risk reduces the number of somatic alterations required for tumorigenesis. These findings support further clinical studies of polygenic risk scores as inexpensive and minimally invasive adjuncts to standard risk stratification. Further studies are required to interrogate generalizability to more ancestrally and clinically diverse populations.

Systematic Assessment of Tumor Purity and Its Clinical Implications.

PURPOSE: The tumor microenvironment is complex, comprising heterogeneous cellular populations. As molecular profiles are frequently generated using bulk tissue sections, they represent an admixture of multiple cell types (including immune, stromal, and cancer cells) interacting with each other. Therefore, these molecular profiles are confounded by signals emanating from many cell types. Accurate assessment of residual cancer cell fraction is crucial for parameterization and interpretation of genomic analyses, as well as for accurately interpreting the clinical properties of the tumor. MATERIALS AND METHODS: To benchmark cancer cell fraction estimation methods, 10 estimators were applied to a clinical cohort of 333 patients with prostate cancer. These methods include gold-standard multiobserver pathology estimates, as well as estimates inferred from genome, epigenome, and transcriptome data. In addition, two methods based on genomic and transcriptomic profiles were used to quantify tumor purity in 4,497 tumors across 12 cancer types. Bulk mRNA and microRNA profiles were subject to in silico deconvolution to estimate cancer cell-specific mRNA and microRNA profiles. RESULTS: We present a systematic comparison of 10 tumor purity estimation methods on a cohort of 333 prostate tumors. We quantify variation among purity estimation methods and demonstrate how this influences interpretation of clinico-genomic analyses. Our data show poor concordance between pathologic and molecular purity estimates, necessitating caution when interpreting molecular results. Limited concordance between DNA- and mRNA-derived purity estimates remained a general pan-cancer phenomenon when tested in an additional 4,497 tumors spanning 12 cancer types. CONCLUSION: The choice of tumor purity estimation method may have a profound impact on the interpretation of genomic assays. Taken together, these data highlight the need for improved assessment of tumor purity and quantitation of its influences on the molecular hallmarks of cancers.

Integrative pathway enrichment analysis of multivariate omics data.

Multi-omics datasets represent distinct aspects of the central dogma of molecular biology. Such high-dimensional molecular profiles pose challenges to data interpretation and hypothesis generation. ActivePathways is an integrative method that discovers significantly enriched pathways across multiple datasets using statistical data fusion, rationalizes contributing evidence and highlights associated genes. As part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancers across 38 tumor types, we integrated genes with coding and non-coding mutations and revealed frequently mutated pathways and additional cancer genes with infrequent mutations. We also analyzed prognostic molecular pathways by integrating genomic and transcriptomic features of 1780 breast cancers and highlighted associations with immune response and anti-apoptotic signaling. Integration of ChIP-seq and RNA-seq data for master regulators of the Hippo pathway across normal human tissues identified processes of tissue regeneration and stem cell regulation. ActivePathways is a versatile method that improves systems-level understanding of cellular organization in health and disease through integration of multiple molecular datasets and pathway annotations.

Landscape of transcriptomic interactions between breast cancer and its microenvironment.

Solid tumours comprise mixtures of tumour cells (TCs) and tumour-adjacent cells (TACs), and the intricate interconnections between these diverse populations shape the tumour's microenvironment. Despite this complexity, clinical genomic profiling is typically performed from bulk samples, without distinguishing TCs from TACs. To better understand TC-TAC interactions, we computationally distinguish their transcriptomes in 1780 primary breast tumours. We show that TC and TAC mRNA abundances are divergently associated with clinical phenotypes, including tumour subtypes and patient survival. These differences reflect distinct responses of TCs and TACs to specific somatic driver mutations, particularly TP53. These data further elucidate how the molecular interplay between breast tumours and their microenvironment drives aggressive tumour phenotypes.

The Proteogenomic Landscape of Curable Prostate Cancer.

DNA sequencing has identified recurrent mutations that drive the aggressiveness of prostate cancers. Surprisingly, the influence of genomic, epigenomic, and transcriptomic dysregulation on the tumor proteome remains poorly understood. We profiled the genomes, epigenomes, transcriptomes, and proteomes of 76 localized, intermediate-risk prostate cancers. We discovered that the genomic subtypes of prostate cancer converge on five proteomic subtypes, with distinct clinical trajectories. ETS fusions, the most common alteration in prostate tumors, affect different genes and pathways in the proteome and transcriptome. Globally, mRNA abundance changes explain only ∼10% of protein abundance variability. As a result, prognostic biomarkers combining genomic or epigenomic features with proteomic ones significantly outperform biomarkers comprised of a single data type.

Widespread and Functional RNA Circularization in Localized Prostate Cancer.

The cancer transcriptome is remarkably complex, including low-abundance transcripts, many not polyadenylated. To fully characterize the transcriptome of localized prostate cancer, we performed ultra-deep total RNA-seq on 144 tumors with rich clinical annotation. This revealed a linear transcriptomic subtype associated with the aggressive intraductal carcinoma sub-histology and a fusion profile that differentiates localized from metastatic disease. Analysis of back-splicing events showed widespread RNA circularization, with the average tumor expressing 7,232 circular RNAs (circRNAs). The degree of circRNA production was correlated to disease progression in multiple patient cohorts. Loss-of-function screening identified 11.3% of highly abundant circRNAs as essential for cell proliferation; for ∼90% of these, their parental linear transcripts were not essential. Individual circRNAs can have distinct functions, with circCSNK1G3 promoting cell growth by interacting with miR-181. These data advocate for adoption of ultra-deep RNA-seq without poly-A selection to interrogate both linear and circular transcriptomes.

BPG: Seamless, automated and interactive visualization of scientific data.

BACKGROUND: We introduce BPG, a framework for generating publication-quality, highly-customizable plots in the R statistical environment. RESULTS: This open-source package includes multiple methods of displaying high-dimensional datasets and facilitates generation of complex multi-panel figures, making it suitable for complex datasets. A web-based interactive tool allows online figure customization, from which R code can be downloaded for integration with computational pipelines. CONCLUSION: BPG provides a new approach for linking interactive and scripted data visualization and is available at http://labs.oicr.on.ca/boutros-lab/software/bpg or via CRAN at https://cran.r-project.org/web/packages/BoutrosLab.plotting.general.

Prediction of early breast cancer patient survival using ensembles of hypoxia signatures.

BACKGROUND: Biomarkers are a key component of precision medicine. However, full clinical integration of biomarkers has been met with challenges, partly attributed to analytical difficulties. It has been shown that biomarker reproducibility is susceptible to data preprocessing approaches. Here, we systematically evaluated machine-learning ensembles of preprocessing methods as a general strategy to improve biomarker performance for prediction of survival from early breast cancer. RESULTS: We risk stratified breast cancer patients into either low-risk or high-risk groups based on four published hypoxia signatures (Buffa, Winter, Hu, and Sorensen), using 24 different preprocessing approaches for microarray normalization. The 24 binary risk profiles determined for each hypoxia signature were combined using a random forest to evaluate the efficacy of a preprocessing ensemble classifier. We demonstrate that the best way of merging preprocessing methods varies from signature to signature, and that there is likely no 'best' preprocessing pipeline that is universal across datasets, highlighting the need to evaluate ensembles of preprocessing algorithms. Further, we developed novel signatures for each preprocessing method and the risk classifications from each were incorporated in a meta-random forest model. Interestingly, the classification of these biomarkers and its ensemble show striking consistency, demonstrating that similar intrinsic biological information are being faithfully represented. As such, these classification patterns further confirm that there is a subset of patients whose prognosis is consistently challenging to predict. CONCLUSIONS: Performance of different prognostic signatures varies with pre-processing method. A simple classifier by unanimous voting of classifications is a reliable way of improving on single preprocessing methods. Future signatures will likely require integration of intrinsic and extrinsic clinico-pathological variables to better predict disease-related outcomes.

The Evolutionary Landscape of Localized Prostate Cancers Drives Clinical Aggression.

The majority of newly diagnosed prostate cancers are slow growing, with a long natural life history. Yet a subset can metastasize with lethal consequences. We reconstructed the phylogenies of 293 localized prostate tumors linked to clinical outcome data. Multiple subclones were detected in 59% of patients, and specific subclonal architectures associate with adverse clinicopathological features. Early tumor development is characterized by point mutations and deletions followed by later subclonal amplifications and changes in trinucleotide mutational signatures. Specific genes are selectively mutated prior to or following subclonal diversification, including MTOR, NKX3-1, and RB1. Patients with low-risk monoclonal tumors rarely relapse after primary therapy (7%), while those with high-risk polyclonal tumors frequently do (61%). The presence of multiple subclones in an index biopsy may be necessary, but not sufficient, for relapse of localized prostate cancer, suggesting that evolution-aware biomarkers should be studied in prospective studies of low-risk tumors suitable for active surveillance.

Mapping genomic and transcriptomic alterations spatially in epithelial cells adjacent to human breast carcinoma.

Almost all genomic studies of breast cancer have focused on well-established tumours because it is technically challenging to study the earliest mutational events occurring in human breast epithelial cells. To address this we created a unique dataset of epithelial samples ductoscopically obtained from ducts leading to breast carcinomas and matched samples from ducts on the opposite side of the nipple. Here, we demonstrate that perturbations in mRNA abundance, with increasing proximity to tumour, cannot be explained by copy number aberrations. Rather, we find a possibility of field cancerization surrounding the primary tumour by constructing a classifier that evaluates where epithelial samples were obtained relative to a tumour (cross-validated micro-averaged AUC = 0.74). We implement a spectral co-clustering algorithm to define biclusters. Relating to over-represented bicluster pathways, we further validate two genes with tissue microarrays and in vitro experiments. We highlight evidence suggesting that bicluster perturbation occurs early in tumour development.

Genomic hallmarks of localized, non-indolent prostate cancer.

Prostate tumours are highly variable in their response to therapies, but clinically available prognostic factors can explain only a fraction of this heterogeneity. Here we analysed 200 whole-genome sequences and 277 additional whole-exome sequences from localized, non-indolent prostate tumours with similar clinical risk profiles, and carried out RNA and methylation analyses in a subset. These tumours had a paucity of clinically actionable single nucleotide variants, unlike those seen in metastatic disease. Rather, a significant proportion of tumours harboured recurrent non-coding aberrations, large-scale genomic rearrangements, and alterations in which an inversion repressed transcription within its boundaries. Local hypermutation events were frequent, and correlated with specific genomic profiles. Numerous molecular aberrations were prognostic for disease recurrence, including several DNA methylation events, and a signature comprised of these aberrations outperformed well-described prognostic biomarkers. We suggest that intensified treatment of genomically aggressive localized prostate cancer may improve cure rates.

VennDiagramWeb: a web application for the generation of highly customizable Venn and Euler diagrams.

BACKGROUND: Visualization of data generated by high-throughput, high-dimensionality experiments is rapidly becoming a rate-limiting step in computational biology. There is an ongoing need to quickly develop high-quality visualizations that can be easily customized or incorporated into automated pipelines. This often requires an interface for manual plot modification, rapid cycles of tweaking visualization parameters, and the generation of graphics code. To facilitate this process for the generation of highly-customizable, high-resolution Venn and Euler diagrams, we introduce VennDiagramWeb: a web application for the widely used VennDiagram R package. VennDiagramWeb is hosted at http://venndiagram.res.oicr.on.ca/ . RESULTS: VennDiagramWeb allows real-time modification of Venn and Euler diagrams, with parameter setting through a web interface and immediate visualization of results. It allows customization of essentially all aspects of figures, but also supports integration into computational pipelines via download of R code. Users can upload data and download figures in a range of formats, and there is exhaustive support documentation. CONCLUSIONS: VennDiagramWeb allows the easy creation of Venn and Euler diagrams for computational biologists, and indeed many other fields. Its ability to support real-time graphics changes that are linked to downloadable code that can be integrated into automated pipelines will greatly facilitate the improved visualization of complex datasets. For application support please contact Paul.Boutros@oicr.on.ca.

Spatial genomic heterogeneity within localized, multifocal prostate cancer.

Herein we provide a detailed molecular analysis of the spatial heterogeneity of clinically localized, multifocal prostate cancer to delineate new oncogenes or tumor suppressors. We initially determined the copy number aberration (CNA) profiles of 74 patients with index tumors of Gleason score 7. Of these, 5 patients were subjected to whole-genome sequencing using DNA quantities achievable in diagnostic biopsies, with detailed spatial sampling of 23 distinct tumor regions to assess intraprostatic heterogeneity in focal genomics. Multifocal tumors are highly heterogeneous for single-nucleotide variants (SNVs), CNAs and genomic rearrangements. We identified and validated a new recurrent amplification of MYCL, which is associated with TP53 deletion and unique profiles of DNA damage and transcriptional dysregulation. Moreover, we demonstrate divergent tumor evolution in multifocal cancer and, in some cases, tumors of independent clonal origin. These data represent the first systematic relation of intraprostatic genomic heterogeneity to predicted clinical outcome and inform the development of novel biomarkers that reflect individual prognosis.

Ensemble analyses improve signatures of tumour hypoxia and reveal inter-platform differences.

BACKGROUND: The reproducibility of transcriptomic biomarkers across datasets remains poor, limiting clinical application. We and others have suggested that this is in-part caused by differential error-structure between datasets, and their incomplete removal by pre-processing algorithms. METHODS: To test this hypothesis, we systematically assessed the effects of pre-processing on biomarker classification using 24 different pre-processing methods and 15 distinct signatures of tumour hypoxia in 10 datasets (2,143 patients). RESULTS: We confirm strong pre-processing effects for all datasets and signatures, and find that these differ between microarray versions. Importantly, exploiting different pre-processing techniques in an ensemble technique improved classification for a majority of signatures. CONCLUSIONS: Assessing biomarkers using an ensemble of pre-processing techniques shows clear value across multiple diseases, datasets and biomarkers. Importantly, ensemble classification improves biomarkers with initially good results but does not result in spuriously improved performance for poor biomarkers. While further research is required, this approach has the potential to become a standard for transcriptomic biomarkers.

Validating reference genes within a mouse model system of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) toxicity.

BACKGROUND: Quantitative real-time PCR (qPCR) is the "gold-standard" technique for measuring mRNA abundances. To correctly compare samples and generate biologically valid results, qPCR data usually require comprehensive normalization to account for sample content variation between reactions. The most common normalization approaches use one or more endogenous controls (reference or house-keeping genes) to adjust the measured levels of experimental genes appropriately. Ideal reference genes are those that display minimal variation across experimental conditions, and thus can vary widely across different biological systems. In particular, toxicogenomic studies of transcriptionally-disruptive toxins, like 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), require careful consideration of reference genes. RESULTS: We examined seven candidate reference genes in 199 mice varying in genotype and time/dose of TCDD exposure. We assessed gene-stability in four ways: (1) the variance of the raw Cq values across biological replicates, (2) the fold-change from basal mRNA levels following treatment, (3) the inter- and intra-group stability evaluated using the NormFinder algorithm, (4) the comparative ΔCq method for each candidate gene. Univariate analyses showed Hprt and Eef1a1 are the two most stable individual reference genes. It has been suggested that using multiple genes would produce a more consistent normalization factor; multivariate analysis was performed using NormFinder. In general, stability increased with the number of genes used, but specific gene-combinations synergized. CONCLUSIONS: We have validated seven reference genes for use in analyzing mRNA abundances in mouse models of TCDD toxicity. The use of multiple reference genes increases stability, providing more consistent normalization and more reliable results. The number of reference genes used should be maximized, based on experimental capabilities (platform, sample availability, etc.). Our results show the benefit of validating reference genes using multiple methods prior to generating large biological datasets.