Yeast Colony Morphology With Deep Learning

ISEF Category: Cellular and Molecular Biology

Subcategory: Other  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: Full Year

The Hook

A yeast colony can act like a tiny city. Change the food supply, and the edges, texture, and growth pattern can shift fast. With a cheap microscope and a Raspberry Pi, you can track those changes frame by frame. Then you can ask which genes help colonies adapt to stress.

What Is It?

This project studies how yeast colonies change shape as conditions on the agar change. Think of a colony like a crowd of people spreading across a floor. If food is rich in one place and scarce in another, the colony does not grow in a neat circle. It can form waves, rings, fuzzier edges, or other patterns that reflect how cells respond to their local environment.

You can measure those patterns with time-lapse imaging. A USB microscope takes repeated pictures, and software like Cellpose or StarDist helps you segment the colony, which means separating the colony from the background and sometimes from individual regions inside it. That gives you numbers such as area, perimeter, circularity, and edge roughness. Those numbers turn a pretty picture into data you can compare across conditions.

The genotype part comes from yeast deletion-collection literature. A deletion collection is a set of strains where one gene has been removed in each strain. By matching your pattern data with published gene-function studies, you can connect a visible colony shape to genes involved in nutrient sensing, stress response, or cell wall growth.

Why This Is a Good Topic

This is a strong science fair topic because you can change one nutrient condition at a time and measure a clear visual response. You get real biology, real imaging, and real data analysis in one project. It also connects to a real problem, how cells sense and respond to stress, which matters in fermentation, infection, and basic genetics. You can build a project that grows from simple comparisons to serious computational analysis.

Research Questions

• How does nitrogen limitation change yeast colony circularity over time?
• What is the effect of glucose concentration on colony edge roughness?
• Does nutrient gradient strength change the timing of branching or ring formation?
• To what extent do different yeast strains show distinct morphogenesis patterns under the same agar conditions?
• Which image features best predict the published phenotype of specific deletion strains?
• How does the addition of a second nutrient source change colony area growth rate?

Basic Materials

• USB microscope with adjustable stand
• Raspberry Pi or similar single-board computer
• MicroSD card and power supply
• Petri dishes and sterile agar plates
• Yeast culture or school-approved yeast starter strain
• Basic pipettes and sterile tips
• Marker for plate labeling
• Ruler or calibration slide
• Consistent LED light source
• Computer for image review and analysis
• Gloves, lab coat, and safety supplies.

Advanced Materials

• Incubator with stable temperature control
• Access to yeast deletion strains or a shared strain library
• Sterile hood or clean bench
• Autoclave or approved sterilization access
• Higher-resolution microscope camera
• Motorized stage or repeatable imaging mount
• Reference standards for imaging calibration
• Colony-picking tools and sterile loop set
• Agar media components for defined nutrient gradients
• Fluorescent reporter strains, if available
• Laboratory notebook with strain tracking system.

Software & Tools

• Cellpose: Segments colony regions or cell-like structures from microscope images for quantitative analysis.
• StarDist: Detects object boundaries and helps separate touching features in dense images.
• ImageJ: Measures colony area, perimeter, and shape features from time-lapse images.
• Python: Automates image processing, feature extraction, and statistical comparisons.
• R: Runs statistical tests and makes publication-style plots for your results.

Experiment Steps

1. Define the biological question and choose one colony trait you will measure first, such as edge roughness, circularity, or growth rate.
2. Select a small set of nutrient conditions that create a clear gradient and are realistic for repeat testing.
3. Plan an imaging setup that keeps lighting, focus, and plate position consistent across the whole experiment.
4. Decide how you will turn each image into numbers, including what counts as background, colony edge, and usable time point.
5. Match your measured traits with published deletion-strain findings, and choose a comparison method that tests for pattern similarity.
6. Build a validation plan with controls so you can tell whether your model or measurements are detecting biology, not camera noise.

Common Pitfalls

• Changing lighting between imaging sessions, which makes colony edges look larger or smaller for reasons that have nothing to do with biology.
• Using plates with uneven agar or moisture, which creates fake nutrient effects and warps colony shape.
• Measuring only final colony size, which misses the morphogenesis changes that make the project interesting.
• Training a segmentation model on too few images, which causes Cellpose or StarDist to fail on new plates.
• Comparing strains without matching age, inoculum density, or media prep, which makes genotype effects impossible to trust.

What Makes This Competitive

A competitive version of this project goes beyond simple before-and-after photos. You would define shape features carefully, validate your segmentation, and test whether those features map onto published gene-function categories. Strong statistical choices matter, especially if you compare multiple strains or multiple nutrient gradients. A polished entry also explains what the model can and cannot predict, which shows real scientific thinking.

Project Variations

• Compare bakery yeast, wild yeast, and lab yeast to see whether different species or strains change colony morphogenesis in the same way.
• Use a colored nutrient gradient or pH gradient to test whether colony shape tracks one stressor more strongly than another.
• Skip strain comparison and focus on one deep-learning pipeline, then benchmark Cellpose versus StarDist for colony segmentation quality.

Learn More

• NCBI PubMed: Search for review articles on yeast colony morphogenesis, nutrient sensing, and deletion collections.
• YeastBook, Genetics: Search for chapters on yeast growth, morphology, and genetics through the journal or NCBI links.
• MIT OpenCourseWare: Look for free genetics, molecular biology, and data analysis course materials.
• NCBI Gene: Look up genes linked to nutrient sensing, cell wall regulation, and colony development.
• ImageJ Documentation: Find tutorials for thresholding, measurements, and time-lapse image analysis on the official ImageJ site.