Dagstuhl Seminar 16261
Integration of Expert Knowledge for Interpretable Models in Biomedical Data Analysis
( Jun 26 – Jul 01, 2016 )
(Click in the middle of the image to enlarge)
Permalink
Please use the following short url to reference this page:
https://www.dagstuhl.de/16261
Organizers
- Gyan Bhanot (Rutgers University - Piscataway, US)
- Michael Biehl (University of Groningen, NL)
- Thomas Villmann (Hochschule Mittweida, DE)
- Dietlind Zühlke (Seven Principles AG - Köln, DE)
Contact
- Annette Beyer (for administrative matters)
Schedule
New technologies in various fields of biomedical research have led to a dramatic increase of the amount of electronic data that is available. Not only is the number of patients or amount of disease specific data increasing, but so is the structural complexity of the data, in terms of its dimensionality, multi-modality and inhomogeneity.
As an important example, the sequencing of the human genome ushered in the era of computational biology and translational medicine, with the promise of novel insights into cancer and other diseases, better therapy options, more effective drugs and improved clinical outcome. The development of novel drugs can be directly linked to insights gained as a result of this genomic revolution.
A significant problem, recognized by both the bio-medical and computational community, is the lack of coordination among researchers in these disparate communities, who do not often have opportunities to share ideas and insights to plan and develop effective collaborative research projects.
The aim of the seminar is to bring together researchers who develop, investigate, or apply methods of machine learning and statistics in biomedical data analysis with experts from knowledge representation and integration as well as bio-medical experts with a strong interest in computation and interpretable models.
In order to advance this important field of research, it is vital to facilitate interdisciplinary exchange of ideas and an intense dialogue. The focus of the seminar will be on the development and optimization of methods and processing pipelines, which offer efficient solutions for structured data analysis appropriate for a wide range of bio-medical application scenarios. In particular, the systematic incorporation of domain knowledge into the approaches will be at the center of interest.
The seminar will be centered about six main areas of interest:
- Structured, inhomogeneous and multi-modal data
- Feature selection and biomarker detection
- Diagnosis and classification problems
- Generative models of bio-medical processes
- Visual analytics and data mining
- Big data mining for clinical impact
Emphasis will be on discussion groups formed by the participants with few introductory talks about recent developments and challenges in bio-medical data analysis and knowledge integration. These groups will report their work in plenary sessions as to ensure that their conclusions are disseminated to all participants.
Press Reviews
- Mittweidaer Forscher bei interdisziplinärem Exzellenz-Seminar auf Schloss Dagstuhl
Posted by Prof. Thomas Villmann on July 15, 2016.
The participants were drawn from three distinct disciplines: Biomedical Research, Machine Learning and Visualizations. On the first day, three overview talks on different aspects of bio-medical research were presented, including an overview of omics and clinical data and databases, a summary of current problems in cancer prognosis and metastasis, and steroid metabolomics and its relevance to disease. On the next two days, there were four overview talks on computer science topics, including machine learning, modeling and visualization. Participants also had the opportunity to give shorter presentations of their current research areas and describe open problems, as well as introduce new and relevant datasets and methods. In total, 16 such short talks were presented, covering various areas of biomedical research and computer science. All talks served as starting points for extensive plenary and individual evening and after dinner discussions about the integration of expert knowledge into data analysis and modeling, specifically targeted to cancer informatics. From these discussions, it was clear that there was an urgent need for interactive collaboration to foster successful analysis and interpretation of biomedical data and the success of such collaboration would hinge on active participation from domain experts from biomedical research, data mining and visualization.
Motivated by this conclusion, we identified a joint project in cancer genomics, which would exploit the expertise represented by the seminar participants. On the fourth day, participants discussed the interactive methodology we will follow in the project. Following this, first results obtained by analysis of cancer data from The Cancer Genome Atlas was presented in a joint talk by representatives from all three disciplines (biology, machine learning, visualization). We will extend this project further in the coming months with active participation from the clinicians and computer scientists. The goal of this effort is not just to solve a relevant and outstanding problem in cancer biology but also to work towards publication of our findings in a high-impact journal authored by all participants. To foster this project, we will establish a Wiki, which will serve as a platform for collaboration and communication.
The participants gave feedback on Friday on the organization and content of the seminar. All participants were appreciative of the open, friendly and constructive atmosphere that made learning and insight possible for experts from very diverse disciplines. Getting to know the basic methods used in each field was seen as the perfect starting point for future collaborations. The idea of a joined wiki page as a collaboration platform as well as the already started joined project were highlighted as especially important. Follow-up-meetings of newly formed interdisciplinary teams were initiated and planned e.g. one in Copenhagen. The participants were very enthusiastic about having a further meeting after about a year to discuss results and new directions resulting from the joint project initiated here. Apart from working on a specific project in cancer biology, the goal of the collaboration is to establish a methodology for interactions, disseminate ideas and protocols among the disciplines and establish a common language to foster understanding.
In summary, biologists, both medical and computational experts in the seminar are enthusiastic about joining forces to solve outstanding problems in understanding biological processes. Many of the machine learning methods presented by participants are ready to be applied in real environments such as in clinical use or in research laboratories, after proper technology transfer. Such technology transfer requires targeted funding and agreed upon protocols to ensure adequate resources and necessary quality control, for subsequent release to the community.
The participants felt that influential members in each community should seek opportunities and avenues to urge the appropriate agencies (NIH, NFS, EU Scientific bodies) to establish a targeted program for technology transfer of computational solutions to challenges in the interpretation of biomedical data. Such a program would solicit competitive funding proposals from groups consisting of both biomedical and computational experts, and require products that are rigorously demonstrated on real problems, as well as satisfy appropriate coding and user interface standards, and where appropriate, satisfy requirements of interfacing or integration with existing established systems currently in use by the community.
In medicine the data is treasure
Whose value's beyond any measure
But it is not surprising
That without analysing
Acquisition is meaningless pleasure
(Michael Biehl and Gyan Bhanot)
Creative Commons BY 3.0 Unported license
Gyan Bhanot, Michael Biehl, Thomas Villmann, and Dietlind Zühlke
- Gyan Bhanot (Rutgers University - Piscataway, US) [dblp]
- Michael Biehl (University of Groningen, NL) [dblp]
- Kerstin Bunte (University of Birmingham, GB) [dblp]
- Aguirre de Cubas (Vanderbilt University - Nashville, US)
- Gert-Jan de Vries (Philips Research Lab. - Eindhoven, NL) [dblp]
- Sebastian Doniach (Stanford University, US) [dblp]
- Shridar Ganesan (Rutgers Cancer Inst. of New Jersey - New Brunswick, US) [dblp]
- Tina Geweniger (Hochschule Mittweida, DE) [dblp]
- Gernoth Grunst (Hennef, DE) [dblp]
- Barbara Hammer (Universität Bielefeld, DE) [dblp]
- Marika Kaden (Hochschule Mittweida, DE) [dblp]
- Hossein Khiabanian (Rutgers Cancer Inst. of New Jersey - New Brunswick, US) [dblp]
- Saurabh Laddha (Rutgers Cancer Inst. of New Jersey - New Brunswick, US) [dblp]
- John Aldo Lee (UC Louvain-la-Neuve, BE) [dblp]
- Pietro Lio (University of Cambridge, GB) [dblp]
- Paulo J. Lisboa (John Moores University - Liverpool, GB) [dblp]
- Markus Lux (Universität Bielefeld, DE) [dblp]
- Elke K. Markert (The Beatson Inst. f. Cancer Research - Glasgow, GB)
- John Martens (Erasmus Univ. - Rotterdam, NL) [dblp]
- Thomas Martinetz (Universität Lübeck, DE) [dblp]
- Friedrich Melchert (Fraunhofer Institut - Magdeburg, DE) [dblp]
- Erzsébet Merényi (Rice University - Houston, US) [dblp]
- Klaus Mueller (Stony Brook University, US) [dblp]
- David Nebel (Hochschule Mittweida, DE) [dblp]
- Jeffrey Rathmell (Vanderbilt University School of Medicine, US)
- W. Kimryn Rathmell (Vanderbilt University School of Medicine, US) [dblp]
- Anupama Reddy (Duke University Medical Center - Durham, US) [dblp]
- Timo Ropinski (Universität Ulm, DE) [dblp]
- Joshua T. Taylor (Rice University - Houston, US)
- Peter Tino (University of Birmingham, GB) [dblp]
- Alexei Vazquez (The Beatson Inst. f. Cancer Research - Glasgow, GB) [dblp]
- Thomas Villmann (Hochschule Mittweida, DE) [dblp]
- Gunther Weber (Lawrence Berkeley National Laboratory, US) [dblp]
- Ole Winther (University of Copenhagen, DK) [dblp]
- Thomas Wischgoll (Wright State University - Dayton, US) [dblp]
- Röbbe Wünschiers (Hochschule Mittweida, DE) [dblp]
- Dietlind Zühlke (Seven Principles AG - Köln, DE) [dblp]
Classification
- bioinformatics
- data structures / algorithms / complexity
- soft computing / evolutionary algorithms
Keywords
- biomedical data analysis
- knowledge integration
- expert interactions
- feature selection and dimensionality reduction
- modeling
- data visualization
