Cortical Microcolumn Simulation with Wnt and SHH

ISEF Category: Cellular and Molecular Biology

Subcategory: Neurobiology  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: Full Year

The Hook

Your brain starts with signals that act like a tug-of-war on a map. A few chemical gradients can tell cells where to go, what to become, and how to line up. That makes cortical patterning a real puzzle you can model on a laptop. You can test whether simple rules can recreate the layered, column-like structure seen in organoid data.

What Is It?

This project asks how a few signaling gradients can shape the early cortex. Wnt, SHH, and BMP are chemical messengers. Cells read them like road signs. Wnt often pushes cells toward more dorsal, cortical identities, SHH tends to pattern ventral regions, and BMP antagonism can change how sharply those regions separate. A reaction-diffusion model turns those signals into equations that track how they spread, decay, and interact.

Think of it like mixing paint on a canvas, except the paint can also move, fade, and block other colors. Your model asks whether those simple interactions can generate microcolumns, small repeating clusters of similar cell states. Then you compare your simulated patterns with published single-cell RNA-seq data from human cortical organoids in HCA Brain. scRNA-seq means single-cell RNA sequencing, a method that measures which genes each cell turns on. That gives you a real data check, not just a pretty animation.

Why This Is a Good Topic

This is a strong science fair topic because you can test a clear model, tune real parameters, and compare the output with public data. You do not need to invent a new wet-lab protocol to do meaningful research. You can ask whether one signaling rule, one gradient shape, or one antagonist strength best matches observed cell-state patterns. That makes the project measurable, analytical, and tied to a real problem in developmental neuroscience and organoid biology.

Research Questions

• How does changing the Wnt gradient shape alter the predicted size and spacing of cortical microcolumns?
• What is the effect of SHH gradient strength on the proportion of dorsal versus ventral cell states in the model?
• Does adding BMP antagonism sharpen boundaries between simulated cortical domains?
• To what extent do different diffusion rates change the stability of repeating pattern clusters?
• Which parameter set best matches cell-state frequencies reported in human cortical organoid scRNA-seq data?
• How does the model fit change when you compare early-stage and later-stage organoid datasets?
• To what extent does noise in the signaling gradients change the repeatability of the simulated pattern?

Basic Materials

• Laptop or desktop computer with enough memory for repeated simulations.
• Python installed with NumPy, SciPy, Matplotlib, and pandas.
• Jupyter Notebook or VS Code for writing and running code.
• Spreadsheet software for tracking parameter sets and results.
• Public scRNA-seq datasets from HCA Brain or GEO.
• A notes document for recording assumptions, equations, and model choices.

Advanced Materials

• Access to a faster workstation or cloud computing credits for batch simulations.
• Python with scanpy for single-cell analysis and seaborn for plotting.
• Parameter search tools such as scikit-learn or scipy.optimize.
• Access to raw or processed human cortical organoid scRNA-seq matrices from HCA Brain.
• A version control system such as Git for tracking model changes.
• Optional access to MATLAB or R for cross-checking analyses.

Software & Tools

• Python: Builds the simulation, runs parameter sweeps, and stores output for analysis.
• Jupyter Notebook: Lets you document assumptions, code, plots, and interpretation in one place.
• scanpy: Handles single-cell RNA-seq data cleaning, clustering, and comparison to your model labels.
• ImageJ: Helps you inspect saved pattern images and quantify spatial features if needed.
• Git: Tracks model versions so you can compare changes without losing earlier results.

Experiment Steps

1. Define the biological question you want the model to answer, such as boundary sharpness, cluster spacing, or cell-state balance.
2. Translate the signaling idea into a simple reaction-diffusion system with a small set of adjustable parameters.
3. Choose the pattern metrics you will measure, such as cluster count, spatial regularity, or domain purity.
4. Build a calibration plan that links simulated states to published scRNA-seq cell types from HCA Brain.
5. Design a parameter sweep that changes one gradient feature at a time, then compare the resulting patterns.
6. Set validation rules that tell you when the model matches the real data well enough to count as a good fit.

Common Pitfalls

• Trying to match every gene in the scRNA-seq dataset, which makes the model too messy to interpret.
• Changing several gradient parameters at once, which hides which signal actually caused the pattern shift.
• Treating organoid cell clusters as exact copies of fetal cortex, which can lead to weak biological conclusions.
• Ignoring scale differences between simulation space and cell counts, which breaks the comparison to real data.
• Using qualitative images only, which makes it hard to prove that one parameter set fits better than another.

What Makes This Competitive

A stronger project goes beyond making a nice simulation. You compare several model versions, test whether BMP antagonism really improves the fit, and use a clear metric for pattern quality. You also separate training data from validation data, so you can show the model predicts more than one dataset. A top-tier entry would make a careful case that one signaling rule explains the observed pattern better than the others.

Project Variations

• Use mouse cortical development data instead of human organoid data to test whether the same gradient rules still fit.
• Swap the outcome metric from microcolumn spacing to gene-expression boundary sharpness and compare which metric is more sensitive.
• Add a second comparison layer, such as dorsal-ventral identity scores versus proliferation scores, to see which biological feature the model predicts best.

Learn More

• HCA Brain: Search the Human Cell Atlas portal for brain and cortical organoid single-cell datasets.
• NIH GEO: Search for published single-cell RNA-seq studies on human cortical organoids and development.
• PubMed: Search review articles on Wnt, SHH, BMP signaling, and cortical patterning.
• MIT OpenCourseWare: Find free systems biology and computational biology course materials for modeling and analysis.
• Development journal: Search for review and research articles on cortical development and organoid patterning.