Enzyme Kinetics Identifiability

ISEF Category: Biochemistry

Subcategory: Other  ·  Difficulty: Intermediate  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Two enzyme models can fit the same curve and still tell different stories. That means a nice graph is not enough if the parameters behind it blur together. Your project asks a sharper question, can your data really separate catalase or lactase rates, or are you chasing a good-looking fit?

What Is It?

Enzymes are proteins that speed up reactions. In this project, you measure a real reaction from catalase or lactase, then build an ODE model, which is a set of equations that tracks how amounts change over time. Think of the model like a map of the reaction, where each parameter is a knob that changes the shape of the curve.

Parameter identifiability asks whether your data can tell those knobs apart. If two different parameter sets produce almost the same curve, the model is not identifiable. That is like trying to identify a song from a fuzzy recording, you may hear the tune, but not every note clearly.

Why This Is a Good Topic

This is a strong science fair topic because you can change one variable at a time, measure a clear signal, and compare data to a model. It connects to food testing, enzyme behavior, and assay design, where bad parameter estimates can lead to wrong conclusions. You can learn wet-lab planning, curve fitting, and how to judge whether a model is truly informative.

Research Questions

• How does substrate concentration change the identifiability of catalase rate constants?
• What is the effect of enzyme source, such as potato versus yeast, on the fit quality of a Michaelis-Menten ODE model?
• Does adding replicate trials improve the recovery of kinetic parameters from noisy lactase data?
• To what extent does the choice of signal, such as foam height or color intensity, change parameter uncertainty?
• Which set of initial guesses leads to the most stable parameter estimates in SciPy fitting?
• How does measurement noise change whether the model can distinguish between competing reaction mechanisms?

Basic Materials

• Household hydrogen peroxide or another safe classroom peroxide source.
• Catalase source such as yeast, potato, or another teacher-approved sample.
• Lactase tablets and milk or lactose solution.
• Clear cups, test tubes, or small containers.
• Digital kitchen scale with 0.1 g accuracy.
• Stopwatch or phone timer.
• Smartphone camera and a consistent light source.
• Notebook or spreadsheet for data tables.

Advanced Materials

• UV-visible spectrophotometer or microplate reader.
• Temperature-controlled water bath.
• Micropipettes and sterile tips.
• Analytical balance.
• Purified catalase and lactase standards.
• Buffer reagents for pH control.
• Computer with Python, SciPy, and Jupyter Notebook.
• Replicate reaction vessels or 96-well plates.

Software & Tools

• Python: Fits ODE models, runs parameter sweeps, and checks identifiability.
• SciPy: Solves the reaction equations and optimizes parameters.
• Jupyter Notebook: Keeps code, notes, and plots in one place.
• ImageJ: Measures image-based signal when you turn photos into data.
• LibreOffice Calc: Organizes raw data and makes quick charts without paid software.

Experiment Steps

1. Define one reaction system, one output signal, and one parameter set you want to estimate.
2. Sketch the reaction network, then decide which terms belong in the ODE model and which can stay fixed.
3. Simulate several plausible parameter combinations, then check which outputs separate cleanly and which overlap.
4. Design a measurement plan that keeps the signal in its linear range and supports direct calibration.
5. Plan controls and replicates that can separate enzyme behavior, substrate limits, and noise.
6. Fit the model from multiple starting guesses, then compare confidence intervals or profile likelihoods to judge identifiability.

Common Pitfalls

• Fitting catalase or lactase data with too many free parameters, which creates a perfect curve but no trustworthy rate constants.
• Using a signal that saturates early, which hides the part of the curve that tells the parameters apart.
• Changing enzyme batch, source, or freshness between runs, which makes biological variation look like model structure.
• Skipping calibration for the measured signal, which leaves you with relative scores instead of real rates.
• Accepting one best-fit solution from SciPy without checking alternate starting guesses, which misses parameter non-uniqueness.

What Makes This Competitive

A strong version of this project does more than fit a curve. It asks whether the parameters are identifiable before any data collection, then tests that claim with real and synthetic data. If you compare two enzyme systems, report uncertainty, and show why one assay gives tighter estimates, you move from a class demo to a real research story. The sharp edge comes from the analysis, not just the wet lab.

Project Variations

• Compare catalase from potato and yeast to see which source gives the tightest parameter estimates.
• Switch from foam height to color or oxygen-sensor data to test how the measurement choice changes identifiability.
• Run a synthetic-data study with added noise to see how many replicates you need before the model can recover the true rates.

Learn More

• BRENDA: Search enzyme entries and kinetic constants for catalase, lactase, and related enzymes.
• PubMed: Search review articles on enzyme kinetics, Michaelis-Menten models, and parameter identifiability.
• NCBI Bookshelf: Find free textbook chapters on enzyme mechanisms and biochemical modeling.
• SciPy Documentation: Look up solve_ivp, least_squares, and optimization routines for ODE fitting.
• MIT OpenCourseWare: Search free biochemistry or mathematical modeling lectures for enzyme rate laws.