According to the 2002 edition of the Physicians’ desk reference, cancer therapies are simultaneously some of the most expensive and least effective. The reason for this lack of effectiveness is that cancer is a complex disease with many different causes, many of which require different treatments. The goal of the research in my group is to integrate experimental data together to create computational, systems level models of how cancer initiates and grows. In particular, we study cancers that arise from the misregulation of the sonic hedgehog signaling pathway—a master regulator of embryonic development.
To make these models, we employ a mixture of high throughput experiments and high performance computation. High throughput experiments such as gene chips, protein chips, and microfluidic assays allow us to gather large sets of quantitative biological data. To integrate this data, we use computational methods taken from engineering and machine learning, such as differential equation modeling, and probabilistic network modeling such as Bayesian networks. Using these tools, we can identify connections between our measurements, and make predictions based on these connections.
The impacts of this work are in both pharmacogenomics and drug discovery. Pharmacogenomics is the study of how our genetic sequence influences our response to drugs. By creating a systems level model of cancer growth, we can predict the effect of a drug given a specific genetic background. For drug discovery, our systems level models allow us to perform what if experiments to predict how well a particular cancer type will respond to a drug before the drug is even developed.