The fields of cancer biology and cell biology are at a turning point. After intensive research over the past 30 years, the genes that give rise to cancer when damaged have been identified, and a new generation of drugs has been designed to target these genes. However, in the first attempts to treat patients with these drugs, a remaining challenge has emerged: the complexity of gene networks. Genes are connected together in intricate networks, so that blocking one gene with a drug affects many other genes, often in unanticipated ways. As a result, even targeted therapeutics can produce unexpected, and often detrimental, results. The complexity of gene networks is beyond the capabilities of human intuition, even when all the genes in the network have been identified. To deal with this challenge, biologists are now turning to the fields of engineering and computer science to build computational models of complex gene networks. Such models help predict the behavior of the gene networks that have gone awry in cancer, and they will someday be used to predict which drugs will be most effective for each individual patient. Thus, cancer and cell biology are entering a new era of integration, where scientists must bring together concepts in basic biology, oncology, and computational sciences to make important advances. Recently, I completed my doctorate work in computational and systems biology at MIT (Massachusetts Institute of Technology), where I studied individual living cells to develop a computational model of a cell death mechanism essential for cancer prevention. Having gained a foundation in both biology and computational modeling, I am proposing in this fellowship to continue to bring these disciplines together - this time to study a biochemical pathway of critical importance to breast cancer: the PI3K pathway, which controls cell growth, proliferation, and death. The PI3K pathway is one of the most commonly mutated pathways in breast cancer and is a major target for new drug development. With the support of this fellowship, I propose to build a computational model of this pathway that directly predicts the response to targeted therapeutics. After I complete this project under the guidance of my mentors at Harvard Medical School and the M. D. Anderson Cancer Center, I plan to continue this path as an independent cancer researcher, bringing together basic cell biology, clinical data from cancer patients, and computational science to develop models of the gene networks that control cancer. Eventually, I will combine multiple pathway models to create an even more comprehensive model of gene networks in cancer, which will tell us which combinations of targeted therapeutics will best benefit each individual patient. To create a model of the PI3K gene network that predicts drug responses, a very large amount of experimental data is required on how various types of breast cancer respond to different types of drugs that target the PI3K pathway. The technologies for performing such large experiments are just now becoming available. I will make use of a technique developed in one of my mentor's labs, reverse phase protein arrays, which enables dozens of genes to be measured simultaneously on thousands of different cells and conditions. I will combine these protein arrays with the latest targeted therapeutics and large libraries of RNAi (a new class of gene inhibitor), which are available to my mentors, to measure the responses of whole gene networks to many different types of inhibition. We are now able to perform these experiments on over 50 different types of breast cancer cells using a newly developed panel of breast cancer cell lines. Such experiments will form the foundation for a computational model that will be constructed using the latest modeling software, also under development in the lab of one of my mentors. The computational model of the PI3K pathway produced in this study has the potential to become an important tool in the treatment of all types of breast cancer because it will predict the most successful treatment for patients who will benefit from PI3-targeted therapy and it will recommend against needless treatment in cases where PI3K-targeted therapeutics will be ineffective. It is reasonable to expect that within 3 years the model will be accurate enough to predict optimal strategies for general classes of breast cancer (for example, luminal vs. basal). With additional calibration, it is possible that the model will predict optimal treatment strategies for individual patients, based on the profiling of the tumor biopsy by protein arrays or other methods. Beyond the clinic, the model will also serve as a research tool for further understanding the PI3K network, and it can be linked to models of other cancer-causing gene networks. Thus, the model developed here will ultimately improve the treatment of breast cancer by helping both clinicians and future researchers to better understand the complex PI3K gene network.