Title: Generalised graphical models for the analysis of phospho-flow cytometry data from drug combination experiments
Authors: Andrea Cremaschi - Universitetet i Oslo (Norway) [presenting]
Manuela Zucknick - University of Oslo (Norway)
Kjetil Tasken - Centre for Molecular Medicine Norway (Norway)
Sigrid Skanland - Centre for Molecular Medicine Norway (Norway)
Abstract: The study of drug interaction via concentration-response experiments has recently received increasing attention. One technique used to produce such data is phospho-flow cytometry, measuring the expressions of a set of proteins of interest, pre-selected according to the study to be undertaken (e.g. proteins targeted by a specific drug or involved in the evolution of a type of cancer). Hence, the phosphorylation level of such proteins can be seen as a real-valued vector of the same dimension as the number of selected proteins. A typical application is the study of drugs that act on cancer cells by stimulating certain signalling pathways. For proteins that are in such pathways, we can expect that their level of activity as expressed by their phosphorylation level will depend on the concentration of the tested drugs. Signalling pathways imply interaction between the proteins of interest and can be depicted as graphs, but knowledge about these structures is based on experiments and typically incomplete. We propose a Bayesian generalised graphical model to study the relation among such proteins, when two compounds are tested simultaneously over a set of pairs of concentrations. In particular, the vectors of phosphorylation levels are modelled via Gaussian graphical models where the graph depends on the concentrations of the two drugs. The proposed framework is applied to a Chronic Lymphocytic Leukemia (CLL) dataset where two drugs used in CLL management are combined.