In our InnovationWell session chaired by Keith Elliston (Co-Founder, President and Chief Executive Officer of Genstruct, Inc.) taking place the morning of 14 October 2008 at Bryn Mawr, we will have the following 3 systems biology perspectives presented:
Darius M. Dziuda (Central Connecticut State University), Ensemble Classifiers and Biomarker Discovery
Frank Tobin (Tobin Consulting), Integrative Mathematical Modeling of Biological Systems
David S. Lester (Innovative Technologies in Health and Wellness, Inc.), Using a Systems Approach to Determine Diabetic Patient Interventions and Outcomes
The perspectives will be followed by a knowledge café discussion, lunch and in the afternoon a further related session on computational biology chaired by Debraj GuhaThakurta (Rosetta Inpharmatics, Merck & Co.):
http://barryhardy.blogs.com/theferryman/2008/08/computational-b.html
A description of the session with presentation abstracts follows. Please add your comments, discussion or questions at the end of the post.
(Please follow continuation here to read abstracts. Comments can be made at the end.)
InnovationWell Systems Biology KM Knowledge Management bioinformatics Computational Biology Translational Medicine Molecular Medicine biology pharmaceutical pharma meeting workshop training Critical Path toxicology Life Sciences Drug Discovery Research and Development Drug Development Healthcare Innovation Innovation Management events
Abstracts
Ensemble Classifiers and Biomarker Discovery
Darius M. Dziuda, Ph.D., Associate Professor of Data Mining & Statistics, Central Connecticut State University
One of the recent trends in supervised learning is the ensemble classifier approach. Classification systems based on multiple individual classifiers often outperform single classifiers, especially when the latter are weak or unstable. An unknown sample is classified by all individual classifiers and assigned to one of the classes in a result of weighted or unweighted voting. However, the ensemble approach generally does not deliver parsimonious biomarkers. On the contrary, ensemble classifiers base their voting on a usually large number of variables represented in all individual classifiers.
One of the main goals of biomarker discovery is to identify a small set of genes (or proteins) whose joint expression profile can significantly separate the differentiated classes and can be used for efficient classification of new cases. Are then ensemble approaches useless for biomarker discovery?
To answer this question, this talk will start with an introduction to ensemble classifiers and then will focus on proper ways of using the ensemble approach in biomarker discovery. Discussed topics will include utilization of the ensemble paradigm in identification of robust and parsimonious biomarkers and in validation of classification systems.
Integrative Mathematical Modeling of Biological Systems
Frank Tobin, Tobin Consulting
Mathematical modeling of complex biological systems is becoming capable of tackling increasingly complicated biology. These models can function as a central, integrative hypothesis to tie together diverse experimental data. Partially this advance is a result of modern systems biology experimentation (i.e. "omics") providing the data needed. There is also a realization that as models increase in size, there is a need to improve the modeling technology itself. This talk will consider a lipid metabolism model as a template for integrative views of modeling not only from the central hypothesis viewpoint (integrating data and phenomena), but also integrating different scientific approaches to scrutinizing the modeling process and to improve upon it. For example, the model is simultaneously a biological explanation and a series of equations with analytical and numerical properties. The model building process for lipid metabolism and the concept of calibration will be discussed from the biological, mathematical, and experimental perspectives. This integration of biology and computation will be considered from the standpoint of hypothesis generation to ask the question - can we learn new biology from the models we've built without being prejudiced by the biological information we put into the model?
Using a Systems Approach to Determine Diabetic Patient Interventions and Outcomes
David S. Lester, Ph.D., President, ITHW, Inc.
Systems dynamics has been used in many industries extending from aerospace to the National Football League. This presentation will demonstrate a new application, developing a model for evaluating the impacts of interventions of the diabetic and prediabetic from an outcomes-based perspective. The reason for the use of systems dynamics is that it is based on the concept that the human brain can only manage three loops in terms of information processing. The factors affecting outcomes for a diabetic patient result in multiple feedback and information loops way beyond the capabilities of normal human brain processing. The development of the model along with the initial prototype will be presented. The model is technology-agnostic, including therapeutics, devices, disease management, diagnostics, etc. Certain assumptions were made including incorporating 10 of the potential 40+ complications and co morbidities in the management of a diabetic patient. All of the data used to develop this model was obtained from publicly available information on the web. It was used to generate rate changes, thus the model provides dynamic outcomes modeling. The potential for the systems dynamics product to make “predictions” or “decisions” will be presented. In addition, further complications will be added in order to demonstrate the levels of complexities and the potentials of the system. It is predicted that up to 750,000+ different categories could be incorporated into this model resulting in what could be considered as “individualized medicine or disease management”. This model demonstrates the diversity of issues that a physician treating the diabetic patient should take into account in terms of managing the diabetic or prediabetic. It provides the framework for a potential decision-making disease management tool that could provide much greater value to the patient if patient outcomes were to be used as the goal for disease intervention and management.
Barry Hardy
InnovationWell Community of Practice & Research Manager
The systems biology approach often involves the development of mechanistic models, such as the reconstruction of dynamic systems from the quantitative properties of their elementary building blocks. Due to the large number of parameters, variables and constraints in cellular networks, numerical and computational techniques are often used.
----------------
hesslei...
http://www.drivenwide.com
Posted by: hesslei | October 14, 2008 at 11:17 AM