Control of biomolecular systems to guide cell phenotypes

About This Grant

Project Summary/Abstract Phenotype control, an active area of research in network control, is applicable to interacting biomolecular systems and can identify targeted interventions that lead to desired cell phenotypes. Phenotype control distinguishes itself from classical control theory in that (i) its objectives are related to dynamical attractors (e.g., stable states), and (ii) its interventions don’t need to be continuously adjusted based on the state of the system. This type of control is well-suited for guided cell differentiation, inducing cell fate changes, or inducing apoptosis of a target cell population. This research program will further develop two phenotype control methods. Feedback vertex set (FVS) control is based on the interaction network that underlies a biological system, and stable motif (SM) control is based on a dynamic model of the system. The PI has participated in the establishment of both of these methods, and has a track record of collaborative construction of experimentally validated dynamic models of biological systems. This research program will overcome the remaining challenge to the wide implementation of each phenotype control method. The barrier to wide application of FVS control is that in many systems the characterization of the target cell phenotype (e.g., the known state of a few biomarkers) is not sufficient to specify the desired state of all FVS nodes. This barrier will be eliminated by identifying the most parsimonious and sufficiently discerning characterization of each phenotype and extrapolating the existing biological knowledge to achieve this characterization. The bottleneck to the wide application of SM control is the long time needed for the development and verification of dynamic models. Automating the key steps of model construction and refinement, building on the recently developed BOOLean MOdel REfiner (boolmore) tool, will substantially decrease this time. The FVS and SM control methods will be implemented on networks constructed from curated pathway databases, on ensembles of dynamic models generated for each network, as well as on select dynamic models constructed in collaboration with domain experts. Integration of the two methods will identify multiple high-confidence interventions for each system. The predicted interventions will be validated via experiments done by the PI’s collaborators (see letters of Prof. Nobile and Prof. O’Rourke). The outcomes of the next five years will include a publicly shared suite of computational methods for efficient control of cell phenotypes as well as novel control protocols for multiple specific systems.