CONICET | Buscador de Institutos y Recursos Humanos

The phenotypes manifested by genetic alleles are influenced by the genetic background in which they reside. Yet, we still have a very limited understanding of how genetic interactions (GIs) influence animal development. The goal of our project is to use genome-wide screens to identify all enhancing and suppressing GIs for a set of strains harboring temperature sensitive (ts) mutations in 24 essential embryonic genes. We have completed over three million primary GI assays and secondary screening of putative suppressors, and we have archived in a database all experimental metadata and images, along with quantitative scoring results from an automated phenotypic scoring algorithm we developed (DevStaR). DevStaR combines computer vision and machine learning methods to count different developmental stages in mixed populations of animals. Using these results we have developed a quantitative phenotypic ?GI score? based on the multiplicative model of independence: if the effects of perturbing two genes are independent, then their combined effects should not deviate from the product of their individual effects. GI scores for individual experimental replicates correlate positively with semi-quantitative manual estimates of interaction strength. Using manual inspection as a reference, we devised criteria to combine GI scores across replicates that reliably detect suppressing interactions. We then generated final interaction scores that reflect both strength and reproducibility, which we used to define ~800 high-confidence and ~750 intermediate-confidence suppressing interactions. Based on comparisons with manual scoring, we estimate the false discovery rates in these two sets as 2% and 10%, respectively. The resulting GI network provides the first genome-wide map of suppressing genetic interactions for the embryo based on quantitative phenotypic analysis of viability.