Recent advances in experimental biology have made available vast amounts of quantitative data describing the complex systems under study; unfortunately there exists a bottleneck in extracting biologically-revelant information from these data. We present here methods for inferring such information from biological image and network data, in which probabilistic models are proposed and statistical inference is used to calculate the parameters and complexity of these models from observed data.

We first present an automated scheme to segment cells in micrograph data and infer leading-edge velocities from time-lapse micrograph sequences. We then show applications of this scheme to several high-throughput studies of cell motility which provide key insight into biophysical models of cell locomotion; results include calculations of spatiotemporal correlations in lateral waves at the cell edge and a fundamental understanding of T cell migration.

We then extend our work on image segmentation to the problem of "segmenting" biological networks, i.e. module discovery. We propose a probabilistic model for modular networks and present a principled, efficient, and interpretable algorithm for approximate Bayesian inference of the model parameters and complexity. We elucidate analogies between module discovery and disorder-averaged spin-glass calculations and show how several existing methods for finding modules can be described as special, variant, or limiting cases of our work. Finally, we apply the technique to synthetic and real networks and outline how the method naturally allows selection among competing network models.