|
Abstract:
|
The yeast two-hybrid assay is a powerful tool for identifying protein-protein interactions, however, only a limited number of proteins can be assayed at a time. Researchers at the University of Iowa have adapted this method to screen and identify multiple protein interactions simultaneously, using a single replicate per gene and next generation sequencing. Using this method, yeast colony counts were obtained for non-binding control vectors and potentially binding proteins, at baseline and again once they had been subjected to an environment of selection pressure for a particular gene. Often in such count data, overdispersion is observed. A Bayesian hierarchical model termed DEEPN was developed to identify true protein-protein interactions. DEEPN accounts for the variability of the count data via estimation of separate overdispersion parameters for the baseline and selected counts. The R package, edgeR, estimates a common overdispersion parameter and allows for a less computationally intensive and time-consuming empirical Bayesian analysis. We compare DEEPN, edgeR, and negative binomial regression for analysis of the DEEPN data.
|
