research Current Research Interests The central theme in my research is the integration of complex data sets and mathematical modeling to understand and engineer cellular signaling systems. My doctoral work focused on understanding signal transduction in human cells – both normal and pathologic – from a systems perspective, using novel computational tools to model, characterize, and predict cellular behavior. This approach was applied to two biological systems: (i) signaling through the epidermal growth factor receptor (EGFR), a receptor tyrosine kinase that is commonly overexpressed or structurally altered in human cancers; and (ii) phosphoinositide and calcium signaling in human platelets. Platelets are small, anucleate cell fragments that respond to vessel injury to prevent blood loss or, under diseased conditions, to initiate thrombosis. Our effort to model EGFR signaling was strongly motivated by the observation that many cancer patients who harbor certain EGFR mutations show a remarkable response to tyrosine kinase inhibitors [1, 2]. Thus, we studied the effects of molecular alterations in the receptor (i.e., mutant forms of the receptor) on its kinetic behavior and downstream signaling responses [3]. By modeling signal flows through branching pathways of the receptor, we showed that EGFR mutants had increased inhibitor binding, enhanced phosphorylation of particular substrate tyrosine residues, and preferential activation of the Akt signaling pathway, a critical pathway for cell growth, survival, and motility. This last prediction is consistent with experimental findings4 and is being pursued further.   In a second project, we developed the first computational human platelet model—assembled from 24 peer-reviewed studies—that accurately predicted the full transient calcium and phosphoinositide dynamics in response to increasing levels of ADP [5]. In a full stochastic simulation of single-platelet response to ADP, the model provided accurate prediction of the statistics of the asynchronous calcium spiking behavior observed in single platelets [6] and provided a quantitative molecular explanation for this stochastic behavior. Interestingly, the asynchronous spiking was a result of the platelet’s small size, suggesting that large populations of platelets may be needed in vivo to achieve a robust signaling response. The analysis also yielded specific, testable predictions regarding the requirement for high SERCA/IP3R ratios in functional platelets, limits on the concentration of intracellular calcium stores, and the relative potency of platelet agonists.   More recently, we have developed a high-throughput protocol to test the platelet response to all pair-wise combinations of 6 agonists at 3 doses. These 15,000 data points representing 154 pair-wise combinations were used to train an artificial neural network (NN) that successfully predicted sequential agonist responses and all ternary combinations of 3 of the platelet agonists. The NN model also identified combinations of 4, 5, and 6 agonists predicted to display synergistic signaling. These predictions are currently being confirmed experimentally. We are also training models for 3 individual normal donors to discover unique patterns of agonist synergisms that will generate a functional fingerprint for each donor. If successful, this approach will ultimately allow us to approximate the conditions in vivo where platelets interact simultaneously with many molecular signals. Future Research Interests   Though my future interests are varied (and open to redirection), there are some common themes that characterize the type of work I would like to pursue during my postdoctoral training and beyond. Broadly speaking, I am interested in the integration of large genomic and biochemical data sets to provide new insights into the mechanistic basis of cellular function. As a systems biologist, I view human cancers as unique examples of aberrant signaling [7] that may be classified for prognostic purposes using computational methods [8]. However, given the heterogeneity of genetic and epigenetic abnormalities involved [9], it will be important to not only classify tumor types but to also identify potential points of intervention within each deregulated pathway. Largely inspired by work to identify oncogenic pathway signatures from gene expression data [8] as well as the elucidation of transcriptional regulation through graph-theoretical models of protein-protein interactions [10], I believe these two techniques may be combined to identify and predict tumor-specific mechanisms of oncogenic signaling. As a logical starting point, one may begin with the simplest case: mutant cell lines that rely on a single gene or pathway that deregulates signaling (“oncogene addiction” [11]). Standard measures of gene expression and transcriptional activation in these cell lines, combined with literature-derived protein-protein interaction networks, may be used to infer the mostly likely pathways [10] that dominate signaling in these cells. From here, one may proceed to more complex cell lines in which two or more independent mutations confer malignant behavior. Ultimately, predictions of key pathway regulators from these analyses [10] may be followed by high-throughput screening assays to identify specific enzyme inhibitors or gene silencing siRNAs [12]. Identifying key regulators of aberrant signaling has already been used to successfully direct rational design of therapies that suppress these signals [13].