Speaker: Jordan Collignon
Title:A High-throughput Pipeline for the Analysis and Annotation of Sectored Yeast Colonies
Abstract: Prion proteins are commonly associated with fatal neurodegenerative diseases in mammals, but are also responsible for a number of harmless heritable phenotypes in the yeast Saccharomyces cerevisiae. In normal conditions, circular yeast colonies exhibit a prion phenotype, displaying a white, pink, or red color related to the fraction of normal (non-prion) protein. While in mammals prion phenotypes are irreversible, in yeast mild experimental manipulations destabilize prion phenotypes, introduce changes in the intracellular prion aggregation dynamics, and cause colonies to exhibit sectors showing both prion (white or pink) and non-prion (red) phenotypes. The precise mechanism of this destabilization and forces influencing the emergence of mixed colony phenotypes are unknown.Images of experimental colonies provide a rich dataset for characterizing the unknown molecular mechanisms influencing destabilization of prion phenotypes and uncovering relationships between colony-level phenotypic transitions, molecular processes, and individual cell behaviors. However, this rich diversity of data is often ignored in practice in traditional biological pipelines, both because colony counting is labor intensive and procedures for characterizing sectored colonies are scarce.Here, I build the first computational pipeline designed for providing a deep analysis of large quantities of sectored yeast colonies in experimental image data. I employ a deep learning strategy that uses synthetic images of yeast colonies as training data to aid in extracting and classifying colonies from real images. At present, this pipeline correctly predicts the frequency of sectors in approximately 91.8% of colonies detected in hand annotated experimental images. Furthermore, this pipeline is able to categorize plates containing more sectored colonies than uniform (fully white or red) colonies, allowing a way to distinguish between plates based on the outcome of experiments on yeast. This approach will streamline quantification and annotation of yeast colonies grown under experimental conditions and offer additional insights into mechanisms driving colony-level phenotypic transitions.