A Novel Computational Method for Inferring Dynamic Genetic Regulatory Trajectories

We present a novel method called Time Series Affinity Propagation (TSAP) for inferring regulatory states and trajectories from time series genomic data. This method builds on the Affinity Propagation method of Frey and Dueck [10]. TSAP incorporates temporal constraints to more accurately model the dynamic nature of underlying biological mechanisms. We first apply TSAP to synthetic data and demonstrate its ability to recover underlying structure that is obscured by noise. We then apply TSAP to real data and demonstrate that it provides insight into the relationship between gene expression and histone posttranslational modifications during motor neuron development. In particular, the trajectories taken by the Hox genes through the space of regulatory states are characterized. Understanding the dynamics of Hox regulation is important because the Hox genes play a fundamental role in the establishment of motor neuron sub-type identity during development [6]. Thesis Supervisor: David K. Gifford Title: Professor of Electrical Engineering and Computer Science