From big-data to bedside


Symposium on translational bioinformatics in cancer research

2nd of April 2015, Ghent

Workshop on 1st of April

Lecture of Lieven Annemans on 8th of September


The Vlaamse Liga tegen Kanker or VLK, a non profit funding agency, recently awarded a lectureship to Katleen De Preter, entitled: "Bioinformatics for future precision medicine in cancer treatment". As a part of this lectureship, a one-day symposium is organized entitled: "From big-data to bedside".

Organizing team from the Center for Medical Genetics Ghent:


Big datasets generated on the newest “omics” platforms significantly help in establishing a comprehensive view and understanding of tumor biology. At the same time, the handling and mining of these data sets poses great bioinformatic challenges. The complexity and size of these datasets is overwhelming and novel bioinformatic tools are being developed at a rapid pace to ensure fast and powerful analyses. More specifically, in relation to translating these big datasets to the bedside, there is a growing belief that precision medicine approaches are required in order to improve selection of patients for particular therapeutic compounds and combination therapies. This symposium will focus on the identification of new targets for therapy and the development of prognostic or predictive biomarkers through analysis of big datasets.

We aim to serve a broad public of basic and translational cancer researchers either with or without expertise in computational biology, interested in an update of the state-of-the art of this rapidly moving but clinically highly relevant research area of precision cancer medicine.


Click titles to see abstracts

08:30 - 09:15Welcome coffee
09:15 - 09:50Katleen De Preter (Ghent University)
View this talk
09:50 - 10:30Aravind Subramanian (Broad Institute, Boston)
Library of Integrated Network-based Cellular Signatures (LINCS)
View this talk
10:30 - 11:00Coffee break
11:00 - 11:40Julio Saez-Rodriguez (EMBL-EBI, Hinxton)
Predictive computational models to fight cancer resistance to therapies
11:40 - 12:20Pieter Peeters (J&J, Beerse)
Big Data in Big Pharma, more than Big Fuss?
View this talk
10:30 - 11:00Sandwich lunch
13:30 - 14:10John Quackenbush (Dana-Farber, Boston)
Solving the Personalized Medicine Puzzle
View this talk
14:10 - 14:50Diether Lambrechts (ERC-VIB, Leuven)
Tumor hypoxia causes DNA epimutations by reducing TET activity
14:50 - 15:30Stefan Pfister (DKFZ, Heidelberg)
Clinical neurooncology in the omics era
15:30 - 16:00Coffee break
16:00 - 16:40Jinghui Zhang (St Jude, Memphis)
Molecular diagnosis for pediatric cancer through integrative analysis of whole-genome, whole-exome and transcriptome sequencing data
16:40 - 17:10Panel discussionView discussion
17:10 - 17:20Closure of the meeting
17:20 - 19:00Drinks

Download full program

What should I bring to the workshop?

09:15 - 09:45Pieter Mestdagh and Steve Lefever
introduction to decodeRNA tool
09:45 - 10:15excercises on decodeRNA
10:15 - 10:45Coffee break
10:45 - 11:30Zeynep Kalender Atak and Hana Imrichova
introduction to iRegulon/i-cis Target
11:30 - 12:30excercises on iRegulon/i-cis Target
12:30 - 13:30Lunch
13:30 - 14:00Corey Flynn and Aravind Subramanian
introduction to cmap/lincscloud
14:00 - 15:00excercises on cmap/lincscloud part 1
15:00 - 15:30Coffee break
15:30 - 16:30excercises on cmap/lincscloud part 2
16:30 - 17:00Jan Koster and Richard Volckmann
introduction to R2 tool
17:00 - 17:45excercises on R2 tool
18:30 - 21:30Meet the expert dinner

Download full program

Katleen will give an overview of the state of the art in translational bioinformatics in cancer research.

The overarching goal of the LINCS program is the development of comprehensive signatures of cellular states that can be used by the entire research community to understand protein function, small-molecule action, physiological states and disease states.

The LINCS effort includes a diversity of cellular readouts of experimental perturbation of which we focus efforts of the LINCS Center for Transcriptomics at the Broad Institute on mRNA expression. Our work began in 2005 to pilot the idea of a ‘functional look-up table’ (the Connectivity Map) whereby researchers could connect signatures of genetic perturbation to signatures of disease states, thereby linking disease physiology to the genome, and connect small-molecules (drugs) to their mechanism of action. Similarly, genes lacking functional annotation could be placed into pathways based on their common perturbational signatures.

While our first foray with the Connectivity Map used Affymetrix arrays, their high cost was prohibitive for large-scale studies. To address this impediment to a true genome-scale Connectivity Map, we validated a reduced representation approach to transcriptional profiling whereby 1,000 transcripts are selected based on their orthogonality, and these measured ‘landmark’ transcripts are then used to computationally infer the remainder of the unmeasured transcriptome.

The Connectivity Map has been widely used as a resource by the research community, with over 21,000 registered users. Our L1000 reduced representation method is an accurate, high throughput and very low cost expression profiling solution that has been extensively validated.

During this presentation we will describe the fundamentals, of Connectivity Map, the L1000 assay, analytical methods and results obtained when these data and methods are brought to bear on therapeutic challenges in biomedicine. In particular, we will describe our efforts to (a) create the world’s most comprehensive resource of perturbational signatures. This will include 2 million L1000 genetic (CRISPER knock-out, shRNA knock-down and ORF overexpression) and small-molecule (drug and tool compound) perturbations spanning 50 cell types of varied lineage and (b) make it possible for biologists and computational scientists worldwide to interact with the data by creating a cloud-based computing infrastructure that combines a highly technical, efficient back-end with modular, user-friendly apps that are designed to facilitate biological discovery.

Aravind Subramanian, Director Computational R&D at the Broad Institute of MIT and Harvard in Cambridge Massachusetts, leads the Connectivity Map group.

As a graduate student in the Whitehead Institute Center for Genome Research, Aravind developed Gene Set Enrichment Analysis (GSEA), a knowledge-based algorithm for the interpretation of high-dimensionality genomic datasets. The methodology is widely used and, in part, drives the Connectivity Map discovery toolkit. In addition, the methodology has been developed to analyze genome-scale pooled shRNA screens for the identification of essential genes in cancer cells (RIGER).

Currently, Aravind leads a team of molecular biologists, software engineers and computational scientists as part of the Connectivity Map effort at the Broad. Their efforts are directed towards developing new technologies for transcriptional profiling and to use these data with pattern recognition approaches to discover relationships between genes, drugs and diseases. The group developed a high-throughput, medium-density, low-cost gene expression-profiling platform, called L1000, which is being used at the Broad and the NIH to massively scale-up the Connectivity Map database to include over 2M perturbational profiles and is being increasingly adopted by industry for therapeutic discovery.

lincscloud team

Predicting the response of a patient to targeted therapies based on molecular markers is a major goal in modern oncology. As a model of the heterogeneous response of patients to treatment, large panels of molecularly characterized cancer cell lines are treated with therapeutic compounds and the efficacy of the compounds is measured. Based on one of these screenings, the Genomics of Drug Sensitivity in Cancer project at the Sanger Institute, we have developed computational models to predict drug sensitivity from genomic features of the cell lines and chemical features of the compounds. In the models, genomic and chemical data is integrated with various sources of prior knowledge whenever possible. Analysis of these models point at molecular processes involved in resistance mechanisms, thus proposing systems to further analyze. These processes are often embedded in signaling networks, which are also the target of many therapeutics. Mathematical mechanistic models can be built based on our knowledge of signaling pathways and phosphoproteomic data upon perturbation with drugs and ligands. These models allow us to analyze biochemically the deregulation of signaling networks in cancer, and how drugs interact with them. This approach can improve our understanding of drug’s mode of action and drug resistance, as well as help us identify novel therapeutic opportunities.

Julio Saez-Rodriguez is a group leader at the European Bioinformatics Institute (EMBL-EBI) since 2010, with a joint appointment in the EMBL Genome Biology Unit in Heidelberg, as well as a senior fellow at Wolfson College (Cambridge).  He is an affiliated member of Sage-Bionetworks and a director of the DREAM initiative to catalyze the development of methods in systems biology. From July 2015 he will be Professor at the Joint Research Center for Computational Biomedicine in Aachen, Germany.

He studied Chemical Engineering at the Universities of Oviedo and Stuttgart, and obtained his PhD at the University of Magdeburg and the Max-Planck-Institute with E. D. Gilles. After this, he was a postdoctoral fellow at Harvard Medical School with Peter Sorger and Doug Lauffenburger at M.I.T., and a Scientific Coordinator of the NIH-NIGMS Cell Decision Process Center.

His group develops and applies computational methods to acquire a functional understanding of signaling networks and their deregulation in disease, and to apply this knowledge to develop novel therapeutics. To this end, his group collaborates closely with experimental groups and pharmaceutical companies.

More about Julio

The life science industries can only envy the predictive ability of the physical sciences that is embodied in computer assisted engineering. Nevertheless, pharmaceutical industries including Janssen and regulatory authorities show a renewed interest in computational simulation for the evaluation of benefits and risks of therapeutic approaches. It still remains a bottleneck that the characteristics of biological components are far less well understood than those of their mechanical counterparts. Theoretical derivation of these is still solidly beyond our reach. However, advances in miniaturization and parallelization of an expanding arsenal of biological readout technologies are beginning to enable the acquisition of sufficient data to learn, generalize and emulate the behavior of biological subsystems. Scaled up Machine Learning approaches are proving instrumental in this endeavor. The methods are currently being deployed in Janssen to learn and predict the biochemical impact of compounds based on their chemical structure or phenotypic effects, or the sensitivities to compounds dictated by a given genetic makeup of cancer cells.

The explosion of primary data processing needs for high volume data sources like massively parallel sequencing and high content imaging, and the leap in scale of Machine Learning approaches translate to increased needs in computational performance. To rise to these challenges, Intel and the micro and nanoelectronics research center Imec, have teamed up with Janssen and the five universities in Flanders, Belgium to found the ExaScience Life Lab, which states the creation of novel supercomputer solutions for applied life sciences as its mission. The talk will cover how at Janssen R&D we make use of high dimensional biology data in combination with computational biology and chemistry to expedite the discovery of safe and effective medicines with the ultimate aim to cure and prevent diseases such as cancer.

Pieter Peeters is heading the European computational biology team in Discovery Sciences at Janssen Research & Development. His group is aiding the drug discovery and development teams in Janssen R&D in their search for novel safe and effective drugs by applying both computational as well as wet lab ‘omics approaches, including chemogenomics, functional genomics and chemical genomics.

He joined Johnson & Johnson in 1999 working on enteric nervous system diseases such as irritable bowel syndrome unraveling some of the molecular mechanism underlying the disease. This work led to several drug discovery projects addressing these new mechanisms. In 2007 he became head of the Functional Genomics department supporting research in the area of central nervous system diseases, metabolic syndrome, and oncology. The responsibilities of his group gradually expanded to also include chemogenomics and high content imaging based approaches aiming at comprehensive understanding of drug action.

Pieter obtained his Ph.D. in Applied Biological Sciences (medical molecular biology) at the Center for Human Genetics from the Catholic University Leuven, Belgium and graduated as a bioengineering (gene and cell technology and interface chemistry) at the same university. During his PhD he studied the role of the ETS-variant gene 6 (ETV6) in different mechanisms for leukemogenesis. This research into the molecular causes of leukemia demonstrated for the first time a role for the JAK2 kinase in the etiology of myeloproliferative neoplasms and leukemias.

Currently, he is the Janssen lead for the ExaScience life lab, a collaborative effort between Intel, IMEC the 5 Flemish Universities and Janssen with the aim of expediting R&D in healthcare by applying high performance computing approaches. In addition he is the Janssen lead for the Innovative Medicine Initiative project on the application of inducible pluripotent stem cells in drug discovery and drug safety testing (StemBANCC).

Since the introduction of second-generation sequencing technologies in 2007, the cost and time required to sequence a genome have fallen dramatically. As DNA sequencing increasingly becomes a commodity, biomedical research is rapidly evolving from a purely laboratory science to an information science in which the winners in the race to cure disease are likely to be those best able to collect, manage, analyze, and interpret data. Here I will provide an overview of the approach we have been developing to deal with the challenge of personal genomic data, including integrative approaches to data analysis and the creation of data portals focused on addressing the most common use cases presented by different user constituencies. By effectively collecting genomic and clinical data and linking information available in the public domain, we have made significant advances in uncovering the cellular networks and pathways that underlie human disease, building predictive models of those networks that may help to direct therapies, and in understanding the distinct requirements of research and medical applications.

Although epigenetic alterations are known to drive oncogenesis, their origin is poorly characterized. Here, we describe how hypoxia, which is pervasive in solid tumors, compromises DNA demethylation, thereby contributing to a global loss of hydroxymethylation in tumors and the hypermethylation of gene promoters. As such, tumor hypoxia drives epigenetic tumor heterogeneity, thus providing a substrate for clonal selection of tumor cells. Consequently, hypoxic tumors from 695 breast and 175 glioblastoma patients display exacerbated hypermethylation signatures when compared to normoxic tumors. Similarly, increasing hypoxia in murine breast tumors exacerbates the hypermethylation phenotype, whereas normalizing the tumor vasculature, thereby reducing hypoxia, rescues it. Overall, this suggests that reversing epimutations by re-oxygenating the tumor may be a novel strategy to defeat cancer.

Background: Recent revolutionary advances in genomics technologies have fostered a large variety of new discoveries in the field of (pediatric) neurooncology, but at the same time pose the option & challenge of applying these new methods in a clinical setting. Accurate classification of some entities at the time of diagnosis and relapses from high-risk entities remain major clinical challenges. To this end, we have developed two programs on a national level addressing these topics, namely Molecular Neuropathology 2.0 for the accurate classification of CNS tumors and the INFORM registry study (INdividualized therapy FOr Relapsed Malignancies in Childhood), which is attempting to rapidly generate personalized tumor profiles and identify therapeutic targets in a clinical diagnostic environment for relapse patients.

Methods: In MNP2.0, DNA methylation fingerprints, which are thought to closely reflect the cell of origin, are used to accurately classify brain tumors into biologically and clinically meaningful subgroups. Amongst a total of 9000 analyzed CNS tumor specimens, we have established a reference set of 2200 samples with very good histopathological and clinical annotation covering ~80 different entities and subgroups. This reference is now used for an individual sample as a comparison to identify the class with the best fit. A web interface to make this reference dataset available to the community is currently being built. The INFORM pilot phase assessed the feasibility of integrating rapid molecular profiling in the clinical management of pediatric cancer patients with progressive or relapsed high-risk malignancies. Whole-exome and low-coverage whole-genome sequencing was performed on tumor and normal DNA, complemented with matched tumor RNA sequencing (Illumina HiSeq2500, ‘rapid’ mode). This allowed reliable detection of copy-number changes, point mutations, InDels, fusion genes and deregulated gene expression. Identified alterations were prioritized according to tumor biological relevance and potential as an actionable drug target, with results discussed in a weekly molecular tumor board composed of clinicians, scientists and pharmacists.

Results: First evidence from ~800 diagnostic cases within the MNP2.0 study suggests that in about 10% of cases the histopathological diagnosis will be changed in a way that affects clinical management of the patient. In about an additional 20% of cases, the diagnosis is refined by revealing a meaningful subgroup that cannot be established by conventional neuropathology alone (e.g., molecular subgroup of medulloblastoma or ependymoma). Ongoing round robin experiments with other centers indicate that the methodology is very robust and it is very well feasible to establish this diagnostic pipeline at other centers. In 2015, a pilot study is starting, which will enable all pediatric brain tumor patients across Germany to benefit from this new diagnostic aid. From Oct 2014 to Jan 2015, 57 patients (average age 13 years) were enrolled from 20 centers throughout Germany in the INFORM pilot phase. Entities included: high-grade glioma (n=12), Ewing’s sarcoma (n=11), rhabdomyosarcoma/DSRCT (n=7), medulloblastoma (n=5), ependymoma (n=4), osteosarcoma (n=4), neuroblastoma (n=4), and others (n=9). Tumor tissue was sufficient for DNA analysis of 52 cases and RNA-seq of 47. The average turnaround time from tissue arrival to molecular results was 25 days. Actionable targets with at least ‘borderline’ evidence (according to a prioritization score harmonized with the other major pediatric precision oncology programs across Europe) were identified in 28 patients (49%). Based on the findings, targeted therapeutics were incorporated in the treatment regime of several patients, with anecdotal reports of marked responses.

Conclusion: Nationwide diagnostic and individualized treatment approaches for pediatric cancer patients based on rapid methylation profiling and next-generation sequencing is feasible. Through MNP2.0 we have already analyzed more than 800 CNS tumor samples prospectively and find changes or refinement of the diagnosis in about one third of cases, which seems to be a good justification for the effort.The results of our INFORM pilot phase show that actionable targets can be identified in roughly half of the patients. The INFORM registry study has now opened, aiming for collecting all molecular information and establishing the required infrastructure for a prospective clinical trial on personalized pediatric oncology.

Stefan Pfister was appointed acting head of the Division Pediatric Neurooncology at the German Cancer Research Center (DKFZ) in 2012. Since 2014 he is professor for pediatric neurooncology at the DKFZ and heading the department permanently. Being a pediatrician by training, Pfister received his MD from Tübingen University, and his clinical education at Mannheim and Heidelberg University Hospitals. As a physician-scientist, he completed postdoctoral fellowships with Christopher Rudd at the Dana-Faber Cancer Institute/Harvard Medical School, and with Peter Lichter at the German Cancer Research Center, Division of Molecular Genetics. Pfister´s research focuses on the genetic and epigenetic characterization of childhood brain tumors by applying next-generation profiling methods and subsequently translating novel findings into a clinical context. For his translational neurooncology projects, Pfister received amongst others the German Cancer Award in 2012.

Pediatric Neurooncology Center Heidelberg

The characterization of the landscape of genetic lesions that underlie cancer has been significantly advanced with the recent application of next-generation sequencing (NGS) technology. This methodology can be used to sequence selected subsets of genes, the whole exome, the whole genome, or the expressed transcriptome (RNASeq). By analyzing tumor with matched normal tissue sample, one should be able to identify almost all somatic and germline lesions within the cancer. To define the genomic landscapes and incidence of germline susceptibility of 21 different subtypes of brain tumors, solid tumors and leukemias, we analyzed >1,000 pediatric cancers and matched control tissue (>2000 total genomes) by whole-genome, whole-exome or RNASeq as part of the St Jude Children’s Research Hospital – Washington University Pediatric Cancer Genome Project (PCGP). Novel computational methods have been developed for detecting single-nucleotide variation (SNV), small insertion/deletion (indel), copy number alteration (CNA) and structural variation (SV) at a high accuracy and sensitivity. An integrative analysis pipeline for Whole-Genome, Whole-Exome and RNASeq data has been developed in a clinical setting for assessing its power in detecting somatic and germline lesions previously characterized by molecular diagnostic assays including Sanger sequencing, fluorescence in situ hybridization (FISH), cytogenetic analysis, mass spectrometric genotyping, immunohistochemistry, fragment analysis and gene expression profiling. A pilot study involving 78 pediatric cancer patients shown 95%, 100%, and 98% sensitivity in detecting somatic lesions, germline lesions and gene fusions previously characterized at a molecular diagnosis lab. Extensive validation of 3,000 somatic lesions from 38 cases shows that the specificity for somatic SNV, indel and structural variation is at 98%, 95% and 84% across the genome. Automated classifiers have been developed using the somatic and germline mutation databases generated by PCGP and other public resources to facilitate assessment of pathogenicity of detected mutations. We show examples of integrative analysis in characterizing intra-tumor heterogeneity and clonal evolution trajectory from diagnosis to relapse in pediatric ALL, rhabodomyosarcoma and osterosarcoma.

According to health economists, decisions related to access of new health technologies, must give priority to interventions (both preventive and curative) which result in the greatest amount of health for the money that is invested. These choices can be made by means of health economic evaluations. In such evaluations, the net costs of an investment (e.g. a new technology) are calculated in comparison to the current alternative, and the ratio between these net costs and the net health benefits is then assessed.

Personalized (or perhaps better: stratified) medicine seems to offer benefits for all stakeholders involved in decisions on market access for new treatments. The stratified approach leverages the latest advances in genetics and molecular biology to create better diagnostic tools and associated targeted therapeutics. The obvious expected benefits are more health effects, avoided waste by better targeting therapies, a controllable budget impact, …. Yet, it is argued in this lecture that market access for such therapies can even be more problematic because of different issues related to the joint development of new therapies and associated biomarker(s), as well as to the performance of test-treatment combinations, and uncoordinated decisions on reimbursement.

Source: Ian Jacob, Ahmad Hussein Awada, Katherine Payne, Lieven Annemans. Stratified Medicine: a call for action. Expert Review in Health Economics


Virginie Lovelingbuilding (VAC Gent)

Koningin Maria Hendrikaplein 70 - 9000 Gent

By public transport

The Virginie Lovelingbuilding is located next to the train station ‘Gent-Sint-Pieters’. Leave the station through the main entrace, the building can be found on your left above the bus and tram stops.

By car

You can park your car in the 'Gent-Sint-Pieters' car park (Sint Denijslaan 244). The Virginie Lovelingbuilding is located next to the car park.

Ghent has a very nice historical center, worth a visit!

"Ghent is Europe's best kept secret", according to Lonely Planet.

Visit Gent

Lonely Planet




Meet the expert dinner