2005 Scientific Session Abstracts
Integrating Breast Cancer Recurrence Models with OncoTCap
Roger S. Day, ScD , (day@upci.pitt.edu); Department of Biostatistics, University of Pittsburgh, Pittsburgh, PA William E. Shirey, University of Pittsburgh; Michele Morris, University of Pittsburgh
Context : Recently, the National Cancer Institute launched the Cancer Bioinformatics Grid (caBIG) initiative, an infrastructure to integrate many cancer data sources and sophisticated data analysis tools. There followed the Integrative Cancer Biology Program, to integrate biomathematical modeling with experimental efforts. The two initiatives will eventually jointly supercharge the work of diverse cancer researchers in basic, translational, and clinical science. Our team explores challenges along this journey.
Technology : The Java program OncoTCap (Oncology Thinking Cap) integrates knowledge acquisition with model-building and model-validation. Uses include:
- Assisting experimental design,
- Developing and validating biological interpretations of observations,
- Providing a highly versatile professional training environment.
Application-builders create and use template forms for statements. The model representation schema is agent-based. Alongside the native simulation engine, one can use non-stochastic, non-agent-based, and non-Java-based code.
Design : Day and Dignam (1995) developed models for tamoxifen resistance, based on cancer cell population dynamics. Day, Shackney, Peters (2005) estimated a model of residual tumor burden and regrowth, by nodal status. Paik et al (2004) developed a model for breast cancer recurrence in node-negative patients treated only with tamoxifen (NSABP B20). These independent models span molecular biology, cell population dynamics, and clinical experiment, tracking genetic and epigenetic changes as a tumor grows, evolves and responds to treatment.
Results : Three design principles are demonstrated: incorporating heterogeneous data sources, combining independent models, and using native and non-native computational engines on one agent-based model description.
Conclusion : The design principles implemented in OncoTCap may fill a critical need in bringing caBIG and ICBP resources to the research community, by providing a unifying interface and framework.