Currently, cancer precision medicine is based on matching genomic data to targeted drugs. This has revolutionized cancer care, but is not always successful. Our overall vision is that a patient’s genomic data would tailor models of relevant biochemical networks. These models would then be analyzed to make more precise predictions for personalized treatment strategies. This would include important features of drug action, such as on- and off-target effects, potential toxicity, dosing, dynamics and sequence.

Within our general focus of single-cell cancer systems biology and pharmacology, our lab has three interacting research areas (see below illustration). Each combines computational and experimental tools, including differential equation and probabilistic modeling, high-performance computing, live-cell imaging, flow/mass cytometry, and multiple “omics” technologies (genomics, transcriptomics, proteomics).

Biochemical networks in mammalian cells remain only partially mapped, and their noisy behavior emerges as dynamic and stochastic cell phenotype. We combine theory and experiments to better understand such networks and how they control phenotypes. Theory informs sufficient experimental designs, and what is experimentally feasible constrains theory. Improved knowledge yields new biological insight and better models for systems pharmacology applications.

Cancer is a multi-variate disease that is unique for each patient. Monotherapies are seldom efficacious. We build models that account for typical experimental data from patients, biologically-relevant phenomena, and anti-cancer drug mechanisms of action. Our aim is to use such models to better predict patient-specific drug regimens, as well as streamline drug development by better predicting responders and potential toxicity earlier in the pipeline. Such model-based prediction is in early stages, but there are few alternatives for solving these complex problems facing clinical and industrial researchers.

Technological innovation often drives generation of new knowledge by making seen what was once invisible. We identify bottlenecks in systems biology and pharmacology applications that guide our methodological research efforts. These efforts facilitate advances in all of our research. Two current experimental technologies under research include using fluorophore combinations for increased multiplexing in immunostaining and live-cell imaging, as well as improving the microwestern array, a high-throughput quantitative western blotting assay. Computationally, we are interested in investigating to what extent GPU computing can accelerate our ordinary differential equation model simulations, as well as how deep machine learning methods can improve analysis of live-cell imaging data.