A fully automated deep learning pipeline for high-throughput colony segmentation and classification.

Area of Science:

• Molecular Biology
• Yeast Genetics
• Bioinformatics

Background:

• Adenine auxotrophy is a widely used genetic marker in yeast research for visualizing genetic and epigenetic events via colony color.
• Manual quantification of yeast colonies is labor-intensive, prone to errors, and lacks reproducibility, hindering large-scale studies.

Purpose of the Study:

• To develop a fully automated pipeline for yeast colony segmentation and classification using neural networks.
• To significantly improve the speed and accuracy of white/red colony counting compared to manual methods.

Main Methods:

• Implementation of cutting-edge neural networks for automated colony segmentation and classification.
• Development of a computational pipeline for processing yeast colony images.
• Utilizing readily available training data for model development.

Main Results:

• The automated pipeline achieved a 100-fold increase in speed for white/red colony quantification compared to manual counting.
• The system demonstrates high accuracy in segmenting and classifying yeast colonies.
• The method is easily integrated into existing experimental protocols.

Conclusions:

• The developed automated pipeline offers a substantial improvement in efficiency and accuracy for yeast genetic screening assays.
• This approach increases the statistical power of experiments utilizing adenine auxotrophy, accelerating research in yeast genetics.