Menstrual Hormone ODE Modeling for PCOS Sampling

ISEF Category: Computational Biology and Bioinformatics

Subcategory: Computational Biomodeling  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

Your hormone levels do not move in straight lines. They rise, fall, and trigger each other like a chain of dominos. That makes menstrual-cycle data perfect for a mechanistic model. If you can predict when a test strip will catch the clearest signal, you can turn messy biology into usable timing.

What Is It?

This project uses a mechanistic ODE model. ODE means ordinary differential equation, a math tool that describes how one value changes over time. Here, the values are hormone levels, such as luteinizing hormone, follicle-stimulating hormone, estradiol, and progesterone. The model tries to recreate the normal menstrual cycle, then adds a PCOS-like perturbation, which means a change in the system that mimics polycystic ovary syndrome patterns.

Think of the cycle like a thermostat with several connected dials. One dial affects the next, and the whole system moves as a feedback loop. Your model can test how changing one parameter shifts the timing and size of hormone peaks. Then you can look for the days when at-home hormone strips would have the best chance of catching a useful signal.

Because you are using public clinical time-series data, you do not need to recruit patients. You can fit the model to published measurements, compare healthy and PCOS-like patterns, and ask which sampling schedule gives the cleanest readout. The math does the heavy lifting, but the biology gives the project meaning.

Why This Is a Good Topic

This is a strong science fair topic because you can test real hypotheses with public data and clear math. You are not just drawing a graph, you are modeling a biological feedback system and checking how well it matches observed hormone patterns. The project connects to fertility tracking, cycle health, and better timing for home test strips. A student can learn data cleaning, parameter fitting, simulation, and model validation, which are all useful research skills.

Research Questions

• How does changing estradiol feedback strength alter the timing of the luteinizing hormone peak??
• What is the effect of increased follicle-stimulating hormone baseline on cycle regularity in the model??
• Does adding a PCOS-like ovulation delay shift the optimal sampling window for home hormone strips??
• To what extent do different parameter sets reproduce published healthy-cycle hormone trajectories??
• Which sampling schedule gives the highest chance of detecting a luteinizing hormone surge in healthy versus PCOS-like simulations??
• How does measurement noise change the model's ability to distinguish a normal cycle from a PCOS-like cycle??

Basic Materials

• Laptop or desktop computer with enough memory to run simulations.
• Python installed with scientific libraries such as NumPy, SciPy, Pandas, and Matplotlib.
• Spreadsheet software for tracking papers, parameters, and data sources.
• PubMed access for finding clinical hormone time-series papers.
• NIH or PMC access for full-text articles when available.
• Reference manager such as Zotero for organizing sources.
• Plotting notebook or script editor such as Jupyter Notebook.
• Calculator for quick parameter checks.

Advanced Materials

• Access to a university cluster or high-memory workstation for parameter sweeps.
• Python with ODE solvers, optimization libraries, and uncertainty tools.
• Bayesian inference software such as PyMC or Stan for parameter estimation.
• Version control system such as Git for tracking model changes.
• ImageJ or similar image-analysis software if you digitize plots from published papers.
• Access to institutional journal databases for broader literature review.
• R for cross-checking fits and running statistical comparisons.
• Data extraction tools for time-series digitization from figures.

Software & Tools

• Python: Runs the ODE simulations, parameter sweeps, and plotting for your hormone model.
• Jupyter Notebook: Keeps code, notes, and figures together so you can test ideas fast.
• SciPy: Solves differential equations and helps fit model parameters to data.
• Pandas: Organizes hormone time-series data from public studies into clean tables.
• PubMed: Helps you find clinical studies with hormone measurements and cycle data.

Experiment Steps

1. Define the hormone feedback loop you want to model, and decide which variables belong in the system.
2. Gather public clinical time-series data, and choose studies with enough sampling points to support parameter fitting.
3. Build a baseline healthy-cycle model, and check whether it reproduces the timing and shape of published hormone curves.
4. Add a PCOS-like perturbation, and decide which biological change you want to represent first, such as delayed ovulation or altered feedback sensitivity.
5. Design a sampling analysis, and test which measurement windows capture the strongest differences between normal and perturbed cycles.
6. Compare model predictions against data you did not use for fitting, and measure how well the model generalizes.

Common Pitfalls

• Using a single paper's hormone curve as if it represents every cycle, which makes the model too narrow.
• Fitting too many parameters at once, which creates many solutions that all look correct.
• Mixing units across studies, which breaks the parameter estimates and shifts the simulated peaks.
• Treating plotted figures as exact raw data without checking digitization error, which adds hidden noise.
• Calling the model a diagnosis tool instead of a pattern model, which overstates what the results can prove.

What Makes This Competitive

A competitive version of this project goes past one curve fit. You compare multiple published datasets, test whether the model still works across different cycle lengths, and quantify uncertainty in the parameters. You also make a stronger claim if you identify a sampling strategy that improves detection under noisy, real-world conditions. A good final entry shows careful validation, clear limitations, and a result that helps explain a biological pattern.

Project Variations

• Use urine test-strip hormone targets instead of serum hormones, and compare which analytes give the best timing signal.
• Model adolescent cycle data or irregular-cycle data to see whether the same parameter structure still works.
• Add a sensitivity analysis that ranks which hormone parameters most change the predicted sampling window.

Learn More

• PubMed: Search for review articles and clinical time-series studies on menstrual-cycle hormones and PCOS.
• NIH Office on Women's Health: Read plain-language background on PCOS and menstrual-cycle biology.
• NCBI Bookshelf: Find free textbook chapters on endocrine feedback and reproductive physiology.
• Endocrine Reviews: Search for review articles on hypothalamic-pituitary-ovarian regulation and PCOS modeling.
• MIT OpenCourseWare: Use systems biology and differential equations course materials to review model building and parameter fitting.
• PMC: Search full-text open-access papers with hormone time-series figures you can digitize.