[PSI]-CIC: A Deep-Learning Pipeline for the Annotation of Sectored Saccharomyces cerevisiae Colonies

Area of Science:

• Yeast genetics
• Computational biology
• Prion biology

Background:

• Yeast prion phenotypes, like [PSI], present as color variations.
• Experimental changes can destabilize prions, leading to sectored colony phenotypes.
• Analyzing sector size and frequency is crucial for understanding prion curing, but current methods are limited.

Purpose of the Study:

• To develop the first computational pipeline, [PSI]-CIC, for identifying and characterizing sectored yeast colonies.
• To enable quantitative analysis of sectored colony features, overcoming limitations in manual data extraction.
• To provide insights into the molecular mechanisms driving prion phenotype curing.

Main Methods:

• Developed a novel neural network architecture trained on synthetic yeast colony images.
• Applied the trained model to analyze real images of [PSI]+, [PSI]-, and sectored yeast colonies.
• Validated the pipeline's performance against hand-annotated experimental images.

Main Results:

• The [PSI]-CIC pipeline accurately predicts the state of approximately 95% of detected colonies.
• It correctly determines the frequency of sectors in approximately 89.5% of detected colonies.
• The pipeline provides a rich set of quantitative metrics for sectored yeast colonies.

Conclusions:

• [PSI]-CIC offers a streamlined approach for analyzing sectored yeast colonies, crucial for prion research.
• The pipeline's ability to quantify sector features provides new insights into prion curing mechanisms.
• Future applications may include categorizing colonies under various experimental conditions for detailed comparisons.