Modeling of signal-response cascades using decision tree analysis

Motivation: Signal transduction cascades governing cell functional responses to stimulatory cues play crucial roles in cell regulatory systems and represent promising therapeutic targets for complex human diseases. However, mathematical analysis of how cell responses are governed by signaling activities is challenging due to their multivariate and non-linear nature. Diverse computational methods are potentially available, but most are ineffective for protein-level data that is limited in extent and replication. Results: We apply a decision tree approach to analyze the relationship of cell functional response to signaling activity across a spectrum of stimulatory cues. As a specific example, we studied five intracellular signals influencing fibroblast migration under eight conditions: four substratum fibronectin levels and presence versus absence of epidermal growth factor. We propose techniques for preprocessing and extending the experimental measurement set via interpolative modeling in order to gain statistical reliability. For this specific case study, our approach has 70% overall classification accuracy and the decision tree model reveals insights concerning the combined roles of the various signaling activities in governing cell migration speed. We conclude that decision tree methodology may facilitate elucidation of signal-response cascade relationships and generate experimentally testable predictions, which can be used as directions for future experiments.