Dr. Madigan’s research involves developing combinatorial personalized therapy for childhood leukemia. This project involves a multidisciplinary team of scientists. Drs. Catherine Madigan and William Roberts are involved in clinical research on childhood leukemias at Rady Children’s Hospital and UCSD. Other investigators span a wide combination of disciplines, such as control theory, artificial intelligence, theoretical physics, statistical mechanics and optimization methods, computational modeling of metabolic and signaling networks, disease models and therapeutic interventions. More specifically, Andrew McCulloch (UCSD Bioengineering) has relevant expertise in bioengineering and systems biology, Giovanni Paternostro (Burnham Institute for Medical Research) in biomedicine and systems biology, Philip Duxbury (Michigan State University) in statistical mechanics and control and optimization in physics, Carlo Piermarocchi (Michigan State University) in quantum control and optimization, Jorge Cortes (UCSD Engineering) in control and optimization in engineering. Additionally we have recently added a collaborator with vast expertise in statistical analysis of clinical and omics data, Nicholas J. Schork, Director of Biostatistics and Bioinformatics, The Scripps Translational Science Institute.
The project aims to solve a key challenge for personalized cancer therapy: the optimization of a defined microenvironment for clinically relevant in vitro testing of drug response using patients’ cells. While promising results have been reported for drug sensitivity studies, techniques using patients’ cancer cells have never reached as wide acceptance as antibiotic sensitivity testing for bacterial pathogens. Demonstration of consistent clinical benefit for antibiotic sensitivity did indeed require the careful standardization of chemically defined media and testing conditions. Recent advances in systems biology and high-throughput technology justify a similar effort for the therapy of cancer.
There is an ongoing effort to predict personalized drug response and toxicity using genomics and other omic information such as transcriptomics or metabolomics, and these genome-wide datasets may also facilitate a more systematic design of in vitro microenvironments. Systems biology models can predict cellular response to cytokines and metabolites, providing the opportunity to identify essential media components. We suggest that this strategy might be only partially valid, because the complexity and variability of biological systems imply that hybrid approaches, which include biological search algorithms in addition to omic data and models, are likely to play an important role in the pharmacological control of living systems. Biological search algorithms are searches that are conducted experimentally, for example with patients’ cells, using algorithmic methods derived from more quantitative sciences and facilitated by high-throughput measurement technology. We and others have described the use of these algorithms to identify optimal drug combinations and very similar methods can be used for other combinatorial biological optimization problems.
This project’s general aim is the optimization and personalization of a defined microenvironment for clinically relevant in vitro testing of drug response using patients’ cells. Optimization of this in vitro microenvironment requires the presence of a clinical gold standard of demonstrated prognostic significance that can be precisely quantified. This is the case for acute lymphoblastic leukemia (ALL) in children, for which recent clinical studies show the prognostic significance of minimal residual disease (MRD) measured 8 and 29 days after the start of chemotherapy. Minimal residual disease can be used to identify a group of patients with comparatively poor outcome. Because leukemia cells can grow in suspension we will focus the microenvironment optimization on the soluble culture medium components.
We will ask the following question: If cancer cells from a patient are undergoing the same standard therapy, for the same number of days (8 days), in different in vitro microenvironments as well as in the patient, which is the microenvironment that yields the therapeutic response closest to that in the patient?
The therapeutic response can be assessed, both in vitro and in the patient, by measuring the number of residual cancer cells after therapy using flow cytometry. In future studies, an optimized and personalized microenvironment will be used to test personalized therapies for each child with leukemia.
The main aims are:
1. Use omic data (including metabolomic and microarray data) from leukemia cancer cells and a systems biology approach to select a set of metabolic substrates and cytokines/growth factors as potential components of a defined medium for in vitro drug sensitivity testing.
2. Optimize the selected combination of factors using a high-throughput experimental strategy and dedicated search algorithms.
3. Personalize the optimized combination of factors using molecular profiles.
4. In parallel novel combinations of drugs will be tested in these cancer cells and will identify the omic data associated with response to specific combinations. We plan to obtain both a clinical relevant in vitro testing strategy and a shortlist of combinations to be tested for each patient. The drug response of non-cancer cells may also be measured in order to estimate the safety of the therapy.
