Optimal Gene Regulatory Network Inference using the Boolean Kalman Filter and Multiple Model Adaptive Estimation

We propose a method for the inference of Boolean gene regulatory networks observed through noise. The algorithm is based on the optimal MMSE state estimator for a Boolean dynamical system, known as the Boolean Kalman filter (BKF). In the presence of partial knowledge about the network, a bank of BKFs representing the candidate models is run in parallel in a framework known as Multiple Model Adaptive Estimation (MMAE). Performance is investigated using a model of the p53-MDM2 negative feedback loop network, as well as application to large numbers of random networks in order to estimate average performance.