The goal of targeted drug development has been to specifically intervene with diseased cells, while leaving healthy cells unharmed. However, the success of targeted therapies has been diminished due to the multiplicity of factors and redundant pathways that dynamically regulate cellular responses. We propose a novel in-silico strategy for patient stratification that may allow the more successful development of targeted therapies.

In our work we show how computational modeling in combination with quantitative experiments can be used for mechanism-based drug design. We illustrate this approach using the ErbB signaling network and data from marketed drugs like Erbitux and Iressa. Based on the mechanism of action of these therapeutics, we combine protein profiling of cell lines with computational modeling to predict the drug activities.

Using k-means clustering we show that receptor expression profiles do not cluster with AKT phosphorylation amongst tumors stemming from the same tissue . Instead, tumor cells of different origin resemble each other with respect to their receptor expression profiles. We predict IC50s for pERK and pAKT in the presence of epidermal growth factor (EGF) , heregulin (HRG) or EGF + HRG in breast and colon cancer cell lines. Furthermore, we discuss whether the activation profile of pERK and pAKT is predictive for proliferation. Based on our findings in cell based assays we present our results on how computational systems biology! models can be used to identify responders to drugs targeting EGF signaling pathway.

This computational systems biology approach to drug design appears to be a very powerful strategy allowing the incorporation of patient stratification into the early stages of the drug development process.