Negative Binomial Modeling of Yeast Two-Hybrid Interactions

To investigate protein-protein interactions, a novel yeast two-hybrid next generation screening (Y2H-NGIS) assay has been developed. Using next generation sequencing it is possible to test thousands of interactions in batch. The assay enriches for interacting preys compared to controls, and a bait control is employed to validate results. Count data from this assay can be modeles usign Negative Binomial distribution. Here I propose two approximations to the Negative Binomial modeling of Y2H-NGIS data, using differential expression analysis and processing count tables from multiple biological replicates. A frequentist and Bayesian approximations are compared and tested in a dataset using barley and powdery mildew effector dataset. The goal is to compare selection and control conditions to identify genuine interactions and minimize false positives inherent in this biological system.