Michael P Schroeder

IntOGen-mutations identifies cancer drivers across tumor types

The IntOGen-mutations platform (http://www.intogen.org/mutations/) summarizes somatic mutations, genes and pathways involved in tumorigenesis. It identifies and visualizes cancer drivers, analyzing 4,623 exomes from 13 cancer sites. It provides support to cancer researchers, aids the identification of drivers across tumor cohorts and helps rank mutations for better clinical decision-making.

In Silico Prescription of Anticancer Drugs to Cohorts of 28 Tumor Types Reveals Targeting Opportunities

Large efforts dedicated to detect somatic alterations across tumor genomes/exomes are expected to produce significant improvements in precision cancer medicine. However, high inter-tumor heterogeneity is a major obstacle to developing and applying therapeutic targeted agents to treat most cancer patients. Here, we offer a comprehensive assessment of the scope of targeted therapeutic agents in a large pan-cancer cohort. We developed an in silico prescription strategy based on identification of the driver alterations in each tumor and their druggability options. Although relatively few tumors are tractable by approved agents following clinical guidelines (5.9%), up to 40.2% could benefit from different repurposing options, and up to 73.3% considering treatments currently under clinical investigation. We also identified 80 therapeutically targetable cancer genes.

Visualizing multidimensional cancer genomics data

Cancer genomics projects employ high-throughput technologies to identify the complete catalog of somatic alterations that characterize the genome, transcriptome and epigenome of cohorts of tumor samples. Examples include projects carried out by the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA). A crucial step in the extraction of knowledge from the data is the exploration by experts of the different alterations, as well as the multiple relationships between them. To that end, the use of intuitive visualization tools that can integrate different types of alterations with clinical data is essential to the field of cancer genomics. Here, we review effective and common visualization techniques for exploring oncogenomics data and discuss a selection of tools that allow researchers to effectively visualize multidimensional oncogenomics datasets. The review covers visualization methods employed by tools such as Circos, Gitools, the Integrative Genomics Viewer, Cytoscape, Savant Genome Browser, StratomeX and platforms such as cBio Cancer Genomics Portal, IntOGen, the UCSC Cancer Genomics Browser, the Regulome Explorer and the Cancer Genome Workbench.

Cancer Genome Interpreter annotates the biological and clinical relevance of tumor alterations

While tumor genome sequencing has become widely available in clinical and research settings, the interpretation of tumor somatic variants remains an important bottleneck. Here we present the Cancer Genome Interpreter, a versatile platform that automates the interpretation of newly sequenced cancer genomes, annotating the potential of alterations detected in tumors to act as drivers and their possible effect on treatment response. The results are organized in different levels of evidence according to current knowledge, which we envision can support a broad range of oncology use cases. The resource is publicly available at http://www.cancergenomeinterpreter.org.

OncodriveROLE classifies cancer driver genes in loss of function and activating mode of action

Motivation: Several computational methods have been developed to identify cancer drivers genes—genes responsible for cancer development upon specific alterations. These alterations can cause the loss of function (LoF) of the gene product, for instance, in tumor suppressors, or increase or change its activity or function, if it is an oncogene. Distinguishing between these two classes is important to understand tumorigenesis in patients and has implications for therapy decision making. Here, we assess the capacity of multiple gene features related to the pattern of genomic alterations across tumors to distinguish between activating and LoF cancer genes, and we present an automated approach to aid the classification of novel cancer drivers according to their role. Result: OncodriveROLE is a machine learning-based approach that classifies driver genes according to their role, using several properties related to the pattern of alterations across tumors. The method shows an accuracy of 0.93 and Matthews correlation coefficient of 0.84 classifying genes in the Cancer Gene Census. The OncodriveROLE classifier, its results when applied to two lists of predicted cancer drivers and TCGA-derived mutation and copy number features used by the classifier are available at http://bg.upf.edu/oncodrive-role. Availability and implementation: The R implementation of the OncodriveROLE classifier is available at http://bg.upf.edu/oncodrive-role. Contact: abel.gonzalez@upf.edu or nuria.lopez@upf.edu Supplementary information: Supplementary data are available at Bioinformatics online.

jHeatmap: an interactive heatmap viewer for the web.

SUMMARY The generation of large volumes of omics data to conduct exploratory studies has become feasible and is now extensively used to gain new insights in life sciences. The effective exploration of the generated data by experts is a crucial step for the successful extraction of knowledge from these datasets. This requires availability of intuitive and interactive visualization tools that can display complex data. Matrix heatmaps are graphical representations frequently used for the description of complex omics data. Here, we present jHeatmap, a web-based tool that allows interactive matrix heatmap visualization and exploration. It is an adaptable javascript library designed to be embedded by means of basic coding skills into web portals to visualize data matrices as interactive and customizable heatmaps. AVAILABILITY jHeatmap is freely available at the GitHub code repository at https://github.com/jheatmap/jheatmap. Working examples and the documentation may be found at http://jheatmap.github.io/jheatmap.

A DNA methylation map of human cancer at single base-pair resolution

Although single base-pair resolution DNA methylation landscapes for embryonic and different somatic cell types provided important insights into epigenetic dynamics and cell-type specificity, such comprehensive profiling is incomplete across human cancer types. This prompted us to perform genome-wide DNA methylation profiling of 22 samples derived from normal tissues and associated neoplasms, including primary tumors and cancer cell lines. Unlike their invariant normal counterparts, cancer samples exhibited highly variable CpG methylation levels in a large proportion of the genome, involving progressive changes during tumor evolution. The whole-genome sequencing results from selected samples were replicated in a large cohort of 1112 primary tumors of various cancer types using genome-scale DNA methylation analysis. Specifically, we determined DNA hypermethylation of promoters and enhancers regulating tumor-suppressor genes, with potential cancer-driving effects. DNA hypermethylation events showed evidence of positive selection, mutual exclusivity and tissue specificity, suggesting their active participation in neoplastic transformation. Our data highlight the extensive changes in DNA methylation that occur in cancer onset, progression and dissemination.

Abstract A1-45: In silico prescription of anticancer drugs to cohorts of 28 tumor types reveals novel targeting opportunities

The development of targeted therapies against altered driver proteins holds the promise of selectively and efficiently eliminating cancer cells. However, high intertumor heterogeneity is a major obstacle to develop and apply therapeutic targeted agents to treat most cancer patients. Here, we present the first large-scale therapeutic landscape of cancer as it stands today in a 6.792 sample cohort covering 28 tumor types. To pursue this goal, we developed a three-step in silico drug prescription strategy. 1) To discover actionable driver events, we first comprehensively identified mutational cancer driver genes by detecting complementary signals of positive selection in the pattern of their mutations across the tumor cohorts. We also identified actionable copy number alteration (CNA) and fusion cancer driver genes. Second, we detected which of these driver genes would have an oncogenic role in the tumor and which ones would lose their function. With these two steps we generated the Drivers Database. 2) Next, we systematically gathered all information available on therapeutic agents; FDA approved and in clinical or pre-clinical stages. We considered three different types of targeting strategies for the cancer driver genes: direct targeting, indirect targeting and gene therapies in clinical trials. Moreover, we designed a set of rules for assigning therapeutic agents to specific genomic alterations beard for the driver genes. By doing this last step, we generated the Drivers Actionability Database. 3) Finally, by combining data of Drivers Database, Drivers Actionability Database and sample data, we developed in silico drug prescription, a novel approach to determine which of the drugs could benefit each of the tumor individuals. In all, in the Driver Database we identified 460 mutational cancer driver genes acting in one or more of the tumor types along with 39 driver genes acting via CNAs or fusions. Fifty of these cancer driver genes are targeted by FDA approved agents, 63 by molecules currently in clinical trials and 74 are bound by pre-clinical ligands. We also identified 81 therapeutically unexploited targetable cancer genes. Lastly, by applying in silico drug prescription we found that only 6.7% of the patients could be treated following clinical guidelines, and were concentrated in only 6 tumor types. Moreover, considering repurposing strategies the fraction of patients that could benefit from FDA approved drugs would increase up to 40%, increasing remarkably the fraction of targetable patients in some tumor types like glioblastoma and thyroid cancer, and up to 72% if considering targeted therapies in clinical trials. In summary, the in silico drug prescription based on Drivers and Drivers Actionability Databases was tested on one of the largest cohorts of tumor samples currently collected for research. The main result highlights the current scope of targeted anti-cancer therapies and its prospects for growth in view of the drugs that are currently in clinical trials or at pre-clinical stages. Additionally, another important output of this work is a ranked list of novel target opportunities for anticancer drug development. Continuous update of drug-target interactions information, and the application of the strategy to larger cohorts, will improve the in silico prescription rules contained within the two databases, thus enhancing its usefulness within personalized cancer medicine. Citation Format: Carlota Rubio-Perez, David Tamborero, Michael P. Schroeder, Albert A. Antolin, Jordi Deu-Pons, Christian Perez-Llamas, Jordi Mestres, Abel Gonzalez-Perez, Nuria Lopez-Bigas. In silico prescription of anticancer drugs to cohorts of 28 tumor types reveals novel targeting opportunities. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A1-45.

Long non-coding RNAs defining major subtypes of B cell precursor acute lymphoblastic leukemia

Recent studies implicated that long non-coding RNAs (lncRNAs) may play a role in the progression and development of acute lymphoblastic leukemia, however, this role is not yet clear. In order to unravel the role of lncRNAs associated with B-cell precursor Acute Lymphoblastic Leukemia (BCP-ALL) subtypes, we performed transcriptome sequencing and DNA methylation array across 82 BCP-ALL samples from three molecular subtypes (DUX4, Ph-like, and Near Haploid or High Hyperdiploidy). Unsupervised clustering of BCP-ALL samples on the basis of their lncRNAs on transcriptome and DNA methylation profiles revealed robust clusters separating three molecular subtypes. Using extensive computational analysis, we developed a comprehensive catalog of 1235 aberrantly dysregulated BCP-ALL subtype-specific lncRNAs with altered expression and methylation patterns from three subtypes of BCP-ALL. By analyzing the co-expression of subtype-specific lncRNAs and protein-coding genes, we inferred key molecular processes in BCP-ALL subtypes. A strong correlation was identified between the DUX4 specific lncRNAs and activation of TGF-β and Hippo signaling pathways. Similarly, Ph-like specific lncRNAs were correlated with genes involved in activation of PI3K-AKT, mTOR, and JAK-STAT signaling pathways. Interestingly, the relapse-specific differentially expressed lncRNAs correlated with the activation of metabolic and signaling pathways. Finally, we showed a set of epigenetically altered lncRNAs facilitating the expression of tumor genes located at their cis location. Overall, our study provides a comprehensive set of novel subtype and relapse-specific lncRNAs in BCP-ALL. Our findings suggest a wide range of molecular pathways are associated with lncRNAs in BCP-ALL subtypes and provide a foundation for functional investigations that could lead to new therapeutic approaches. Author Summary Acute lymphoblastic leukemia is a heterogeneous blood cancer, with multiple molecular subtypes, and with high relapse rate. We are far from the complete understanding of the rationale behind these subtypes and high relapse rate. Long non-coding (lncRNAs) has emerged as a novel class of RNA due to its diverse mechanism in cancer development and progression. LncRNAs does not code for proteins and represent around 70% of human transcripts. Recently, there are a number of studies used lncRNAs expression profile in the classification of various cancers subtypes and displayed their correlation with genomic, epigenetic, pathological and clinical features in diverse cancers. Therefore, lncRNAs can account for heterogeneity and has independent prognostic value in various cancer subtypes. However, lncRNAs defining the molecular subtypes of BCP-ALL are not portrayed yet. Here, we describe a set of relapse and subtype-specific lncRNAs from three major BCP-ALL subtypes and define their potential functions and epigenetic regulation. Our data uncover the diverse mechanism of action of lncRNAs in BCP-ALL subtypes defining how lncRNAs are involved in the pathogenesis of disease and the relevance in the stratification of BCP-ALL subtypes.

SVGMap: configurable image browser for experimental data

SUMMARY Spatial data visualization is very useful to represent biological data and quickly interpret the results. For instance, to show the expression pattern of a gene in different tissues of a fly, an intuitive approach is to draw the fly with the corresponding tissues and color the expression of the gene in each of them. However, the creation of these visual representations may be a burdensome task. Here we present SVGMap, a java application that automatizes the generation of high-quality graphics for singular data items (e.g. genes) and biological conditions. SVGMap contains a browser that allows the user to navigate the different images created and can be used as a web-based results publishing tool. AVAILABILITY SVGMap is freely available as precompiled java package as well as source code at http://bg.upf.edu/svgmap. It requires Java 6 and any recent web browser with JavaScript enabled. The software can be run on Linux, Mac OS X and Windows systems. CONTACT nuria.lopez@upf.edu