31 C.Ryan* D. Greene G. Cagney* P. Cunningham 2010 Inferring epistatic relationships for an improved functional overview of yeast protein complexes VIBE 2010, DCU, Dublin, Ireland <p>Epistatic miniarray profiles (E-MAPs) provide a high-throughput methodology to quantitatively measure the strength of pairwise genetic<br /> interactions. Scores are assigned to interactions based on the divergence in growth of yeast strains with two disrupted genes from the expected growth rate.</p> <p>E-MAP datasets are typically represented in the form of a symmetric matrix, with each entry representing a measured interaction. Existing EMAPs vary in size from 400 - 750 genes, and each covers a specific subset of related genes (e.g. Chromosome Biology, RNA Processing). Focusing on subsets of functionally related genes maximizes the number of epistatic interactions identified, and gives an overview of the function in question, but means that interactions between different cellular functions are not as frequently measured.</p> <p>A considerable number (~ 30 - 160) of genes overlap between different EMAP sets, raising the possibility that the correlation between related genes could be exploited in order to predict epistasis scores not directly measured in an individual E-MAP. Nearest neighbor methods have been used to accurately impute missing values in gene expression datasets. They are effective because genes which are coregulated tend to display coherent expression profiles. With E-MAPs, genes from the same complex or pathway tend to display<br /> similar interaction profiles. Here we develop a nearest neighbor based approach and use it to predict new epistatic interactions that enhance our understanding of the yeast interaction network.</p>

* Non-Clique Members