Plating and imaging of yeast colonies for segmentation and classification with neural networks

The yeast research community commonly makes use of a simple and powerful color reporter assay, based on adenine auxotrophy, to study genetic and epigenetic mechanisms. Manual classification and quantification of colony color in high numbers is tedious, immensely time consuming, can be erroneous and irreproducible. To overcome these problems, we have established an automated pipeline to quantify and classify yeast colonies from images of entire plates with high accuracy (98.6%). Having the same plating and imaging conditions as used for training of the network is key to this high accuracy. Here we provide a standardized protocol for plating of yeast and imaging within 5-7 days, compatible with the classification pipeline described in Carl et al (bioRxiv; doi: https://doi.org/10.1101/801845).


Introduction
Adenine auxotrophy is commonly used in yeast research to monitor genetic and epigenetic events.
Mutation or silencing events interrupting the adenine synthesis pathway lead to the accumulation of a red pigment under adenine limiting conditions. Colonies with a functional adenine synthesis pathway appear white. An advantage of this non-selective reporter system is that a direct semi-quantitative comparison of expression levels between colonies can be made on the same plates. Furthermore, the degree and stability of expression (pink, variegating and red) can be determined by comparing the amount and distribution of accumulated red pigment within a colony. A disadvantage of this nonselective reporter is the very time-consuming and tedious color classification of colonies by hand.
Further, if components of the adenine synthesis pathway are expressed at low level (pink phenotype), classification can vary between researchers and screens, potentially leading to erroneous and not fully reproducible results. To reduce these issues, we have developed a highly accurate and automated colony classification method based on neuronal networks with whole plate images as input (Carl et al;bioRxiv, doi: https://doi.org/10.1101/801845). In this protocol we describe in detail the procedures for plating and imaging as used for the pipeline training, which are crucial to achieve optimal results. 4. If desired, add antibiotics and stir for 2 min to ensure a homogeneous mixture.

Reagents
5. Pure 20 ml of autoclaved agar media into each 10 cm plate.
6. After solidification (c.a. 20 min at room temperature), transfer plates upside down into plastic bags and store in a cold room until use or up to two months.

Note: For consistent results keep autoclaving time the same for all plates within an experiment.
Reduced autoclaving time can lead to slightly increased accumulation of the red pigment.
Note: When working with PMGc plates, low thiamine concentration reduces epigenetic ade6 + silencing phenotypes in S. pombe. 3. Pipette 1 ml onto a cover slip and count cells under the microscope. 4. Dilute to 500 cells / 100 ml (per plate).

Incubation of yeast plates
Incubate plates upside down at 30°C for 5-7 days. 2. Adjust camera height or zoom that the entire plate fits into one image.
Note: If plates were previously stored in the cold, let them warm up to room temperature before imaging to avoid classification artefacts caused by condensation.
Note: Different colored backgrounds can possibly be used, however a black background is recommended.
Note: Set color balance to manual and not automatic to avoid increased red in images with only white colonies.

Running the automatic classification
Run the classification pipeline as described in (Carl et al; bioRxiv, doi: https://doi.org/10.1101/801845): 1. Download and install the required python packages as detailed on the github page.
2. Download the classification pipeline script.
3. Place all images that you would like to classify into the same folder.
4. Run the pipeline as described on the github page.

Few colonies classified:
When colonies are very dense (>500 colonies / plate), segmentation can frequently fail, leading to few classified colonies. Also colonies smaller than 20 x 20 pixels are not segmented. This can be easily resolved by decreasing plating density.

Misclassified colonies:
White colonies can be misclassified as pink, if plates were incubated too long (a ring forms around the center of the colony). Reduce incubation time of plates. To check classification of specific colonies, the classification pipeline can be run in "prediction-only" mode on previously-segmented images.

Overestimation of white colonies:
The accumulation of the red pigment decreases growth. If cells are plated too dense, red colonies do not reach the threshold size for segmentation, leading to a relative underestimation of red colonies in comparison to other colony phenotypes.
Further improvement: