Toward merging image analysis and gene network discovery from fluorescent in situ hybridization data

Cellular response to extracellular stresses can now be quantitatively monitored using fluorescent in situ hybridization (FISH). Deducing a cell's transcriptional response mechanism from FISH data requires counting individual RNA transcripts within cells and subsequently enumerating the modes of gene expression (termed gene states). However, when the number of gene states are unknown standard likelihood based parametric inference methods are insufficient, as they require pre-specifying the number of gene states. Here we explore a nonparametric progression of these likelihood-based approaches to deduce the number of gene states alongside relevant rates. If time allows, we will also discuss our method for providing the initial snap-shot data for gene network inference (which requires quantifying diffraction-limited RNA spots in crowded environments): a Bayesian nonparametric image analysis tool which attempts to realize the theoretical maximum number of RNA transcripts identifiable from multiplexed FISH data.