Quantitative Systems Pharmacology

Cellular behavior emerges from a complex network of chemical interactions, the details of which remain largely unknown. Many pathways within the cell are redundant or interdependent, hindering our ability to experimentally delineate their in vivo activity. Further complicating matters are slight and unmeasurable differences between cells within a population and the interplay between cells and their environment. Thus, seemingly identical cells may respond differently to the same environmental perturbation. This cellular heterogeneity is particularly problematic in cancer, where slight differences determine whether or not a cell will go on to produce a tumor. Our lab has several on-going projects modeling disease emergence and progression from sub-cellular to multi-cellular systems. These include:

Mechanistic Pathway Modeling

Interactions between molecules within a single cell are represented graphically as a network of chemical reactions. Mathematically, the reaction network constitutes a set of coupled differential equations, and the chemical behavior of the cell is found by solving the set of equations. Working with experimental collaborators, we are constructing mechanistic models of STAT3, GPCR and IGFR pathways. The models are executed using mass action kinetics or stochastic algorithms, and their parameters obtained numerically through Monte Carlo techniques. These models provide hypotheses about pathway topologies, as well as predictions on cellular response to specific perturbations.

Multivariate High Content Analysis

The single-target view of drug discovery is yielding to a systems-based view in which drugs are used in combinations to target specific and multiple pathways. This approach relies on cell-based, rather than protein-based, screening. As part of the University of Pittsburgh Drug Discovery Institute, we are forging new methods for lead development based on high content screening (HCA). Here, cells are described in terms of tens to hundreds of features relating to the concentrations and localizations of biomolecules, as well as morphological properties. We are using perturbation theory to develop methods for efficiently selecting sets of non-overlapping and informative features, and we are employing maximum likelihood and function theoretical methods toward describing and comparing heterogeneous cell populations.

Heterogeneity in Growing Tumors

It is now well-established that tumors display a high degree of genetic and non-genetic heterogeneity. Although such tumor heterogeneity may hinder the identification of optimal therapies based on single-site biopsies, it may contain a wealth of information on the nature of tumors themselves. We are modeling tumor growth using stochastic models to gain insight into the origins of tumor heterogeneity. By comparing patterns of heterogeneity in clinical tumor samples to our model results, we hope to identify some of the underlying physical and chemical factors that influence tumor growth.

Multiscale Protein Structural Dynamics

As a class of molecules, proteins exhibit the remarkable ability to be simultaneously stable and sensitive. Folded proteins are thermodynamically stable, but many of them are highly sensitive to specific environmental perturbations. This sensitivity is likely a result of energetic frustration: the physical constraints of the peptide chain force some regions of folded proteins into energetically unfavorable conformations, and the native state has the minimal global energy despite localized high-energy regions. We continue to investigate how protein structure influences intrinsic motions and how local energetic frustration can specifically enhance a the physical sensitivity of proteins. The elastic network model (ENM) has enjoyed considerable success in predicting the large-scale equilibrium dynamics of proteins and their complexes. In the typical ENM, each amino acid residue is represented by a single point, usually coincident with its alpha carbon atom, and nearby residues interact with each other through virtual springs of uniform force constant. As the potential is harmonic, the resulting dynamics can be found through normal mode analysis. The basic ENM theory has been expanded upon through a variety of techniques that permit the study of larger and more complex systems. Although the ENM provides a decent approximation to the energy landscape in the vicinity of the native state, it may still be improved upon with simple modifications that do not impinge upon its elegant simplicity.