Modeling B-Cell Affinity Maturation

ISEF Category: Biomedical and Health Sciences

Subcategory: Immunology  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A vaccine is not just about what you give, it is also about when you give it. Inside a germinal center, B cells compete like students in a tournament, and the spacing between doses can change who gets ahead. If you can model that timing, you can ask which prime-boost interval gives the strongest antibody response.

What Is It?

Your immune system does not make every antibody the same way. After a dose, B cells enter germinal centers, which are tiny training camps in lymph nodes. There, cells with better receptors are more likely to survive and multiply. That process is called affinity maturation, which means the average binding strength of the B cells improves over time.

An agent-based model gives each B cell its own rules, like a game with many players. You can change the schedule of antigen doses, then watch how the simulated population responds. Public serology time-series are published measurements of antibodies in blood over time, and you can use them to check whether your model follows real data. That lets you compare prime-boost intervals without running a wet lab yourself.

Why This Is a Good Topic

This topic works well because you can change one clear variable, the spacing between doses, and measure a clear outcome, predicted affinity or antibody titer. It also connects to a real problem, how vaccine timing shapes immune response, so your results feel meaningful. You can learn modeling, calibration, and data validation from public datasets, which makes the project realistic for a student who is new to research.

Research Questions

• How does changing the interval between the prime and boost affect predicted mean antibody affinity?
• What is the effect of antigen dose size on the fraction of high-affinity B-cell clones in the model?
• Does a longer germinal-center reaction time improve the fit between simulated and public serology curves?
• To what extent does calibration on one public dataset transfer to a second dataset?
• Which dosing schedule, short interval or long interval, produces the highest area under the predicted antibody titer curve?
• How does initial B-cell diversity change the schedule that maximizes affinity maturation?

Basic Materials

• Laptop or desktop computer with a stable internet connection.
• Python version 3.11 with Jupyter Notebook installed.
• Spreadsheet software such as Google Sheets or LibreOffice Calc.
• Public serology time-series datasets in CSV or Excel format.
• Notebook for tracking model assumptions, data sources, and results.

Advanced Materials

• Workstation with more CPU cores for repeated simulation runs.
• R or MATLAB for mixed-model checks and figure polishing.
• Version control setup with Git and GitHub.
• Access to longitudinal antibody datasets from public repositories or collaborators.
• Cloud compute credits for large parameter sweeps.

Software & Tools

• Python: Runs the agent-based simulation and parameter sweeps.
• Jupyter Notebook: Keeps code, notes, and figures in one place.
• R: Fits summary models and makes publication-style plots.
• Google Colab: Lets you run Python without local setup.
• GitHub: Tracks code changes and helps you compare model versions.

Experiment Steps

1. Define the exact outcome you will compare, such as peak affinity, time to peak, or area under the antibody curve.
2. Choose the biological rules your agents will follow, including birth, selection, mutation, and decay.
3. Select public serology datasets that match your question, then decide which set will calibrate the model and which set will test it.
4. Build a comparison plan that turns each schedule into the same summary metrics, so different studies stay comparable.
5. Stress-test the model with sensitivity checks, then see whether your ranking of prime-boost intervals stays the same.

Common Pitfalls

• Mixing antibody, binding, and neutralization datasets in one analysis, which makes the schedule look consistent when the measurements are not the same.
• Calibrating on every available dataset, which leaves no untouched data to prove the model can predict new results.
• Ignoring assay timing, which shifts peaks and can make one dosing interval look better just because samples were collected later.
• Letting dose size and dose spacing change together, which hides whether timing or antigen amount drives the model output.
• Reporting only the best-fit schedule, which hides uncertainty and makes small parameter changes look more certain than they are.

What Makes This Competitive

The stronger version of this project compares several prime-boost schedules against held-out public data, not just one published curve. You also report confidence bands or posterior ranges, then show which biological assumptions flip the answer. That kind of validation and sensitivity work turns the model into a research argument instead of a plot. If you can tie the schedule ranking to a clear mechanism inside the germinal center, your project stands out.

Project Variations

• Model how different vaccine platforms, such as protein, mRNA, or viral vector, change the best prime-boost interval.
• Swap binding titer data for neutralization data and test whether the preferred schedule changes.
• Add age-group parameters to the model and compare whether adolescents, adults, and older adults peak at different times.

Learn More

• NCBI Bookshelf: Free immunology chapters on B cells, germinal centers, and adaptive immunity, found by searching the NCBI Bookshelf site.
• PubMed: Search review articles on germinal-center selection, affinity maturation, and vaccine spacing.
• NIH/NIAID: Background pages on antibody responses and vaccine research, available on the NIAID website.
• CDC Immunization Schedules: Real-world vaccination timing context and dose spacing guidance, available on CDC.gov.
• MIT OpenCourseWare: Free lectures on probability, differential equations, and computational modeling, found on the MIT OpenCourseWare site.