Single-Cell Fibroblast Stress Signatures

ISEF Category: Cellular and Molecular Biology

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

The Hook

Some cells keep working even when tissues are under stress. Fibroblasts are one of those cell types, and they help hold organs together like a repair crew. If you can spot a stress-resilient fibroblast subtype, you can ask a real biology question with big medical meaning.

What Is It?

Single-cell atlases are huge maps of individual cells from many tissues. Instead of averaging all the cells together, scientists can look at each cell on its own and see which genes are turned on. That matters because two cells with the same label, like fibroblast, can still behave very differently.

This project asks whether a shared stress-resilient fibroblast subtype appears across mammals. Think of it like comparing several versions of the same app on different phones. The app may look different on the surface, but the core code can still match. Here, you would compare cells from datasets like Tabula Sapiens and Tabula Muris, then look for a gene pattern that stays stable across species and tissues.

ML means machine learning, which is a way to let a computer find patterns in large data sets. In this project, ML would help you build a gene signature, a small set of genes that predicts which fibroblasts belong to the same stress-related state. The key question is not just whether the cells look similar, but whether a model can separate them using real biology instead of noise.

Why This Is a Good Topic

This is a strong science fair topic because the data already exist, but the analysis is still open-ended. You can ask a narrow question, test it with real datasets, and compare methods instead of just repeating a textbook result. The project connects to wound healing, fibrosis, aging, and tissue repair, so the stakes are real. You can also learn single-cell analysis, feature selection, and model validation, which are useful skills for future research.

Research Questions

• How does integrating Tabula Sapiens and Tabula Muris change the separation of fibroblast states across tissues?
• What is the effect of feature selection method on the stability of a fibroblast stress-resilience gene signature?
• Does a gene signature trained on one mammal transfer to fibroblasts in another mammal?
• To what extent do stress-response genes overlap with fibroblast clusters labeled by tissue repair or matrix remodeling markers?
• Which minimal gene set best predicts a conserved stress-resilient fibroblast subtype across datasets?
• How does batch correction method affect whether the same fibroblast subtype appears conserved across species?

Basic Materials

• Computer with at least 16 GB RAM.
• Internet access for downloading public single-cell datasets.
• Tabula Sapiens dataset access.
• Tabula Muris dataset access.
• Spreadsheet software for tracking samples and labels.
• Python or R installed locally.
• Jupyter Notebook or RStudio.
• External hard drive or cloud storage for backups.

Advanced Materials

• Access to a university workstation or server with high memory.
• Public single-cell RNA-seq matrices from Tabula Sapiens and Tabula Muris.
• Curated fibroblast marker gene lists from peer-reviewed papers.
• Reference genome or gene ortholog mapping files.
• Git for version control.
• Conda or another package manager.
• High-memory plotting and analysis environment for integration and validation.

Software & Tools

• Python: Runs data cleaning, clustering, and machine learning workflows for single-cell analysis.
• R: Handles statistical tests, visualization, and single-cell workflows with Bioconductor tools.
• Scanpy: Supports preprocessing, clustering, and embedding of single-cell RNA-seq data.
• Seurat: Helps integrate datasets and compare cell populations across studies.
• GitHub: Stores code, tracks changes, and makes your analysis easier to reproduce.

Experiment Steps

1. Define the exact fibroblast populations and stress markers you will compare across species and tissues.
2. Choose how you will map genes between mammals so your comparisons stay biologically fair.
3. Select one batch correction or integration method and one baseline method for comparison.
4. Decide how you will turn gene expression into a classifier or signature score.
5. Plan validation rules that test whether the signature generalizes to a held-out dataset.
6. Set up figures and summary statistics that show both biological meaning and model performance.

Common Pitfalls

• Comparing gene names across species without ortholog mapping, which can make the same cell look different for the wrong reason.
• Mixing tissues with very different fibroblast roles, which can hide the conserved subtype you are trying to find.
• Letting batch effects dominate the embedding, which can group cells by dataset instead of biology.
• Building a gene signature from all available cells, which can make the model memorize the training data.
• Treating cluster labels as truth without checking marker genes, which can turn a noisy cluster into a fake subtype.

What Makes This Competitive

A stronger project would test whether the same fibroblast state survives multiple integration methods, not just one. It would also separate true conservation from batch effects by using held-out datasets and ortholog-aware validation. The best version would compare model performance across tissues, species, and stress conditions, then explain which genes carry the signal. That kind of analysis shows real control over the data, not just a pretty cluster plot.

Project Variations

• Compare fibroblast stress signatures across healthy, wounded, and fibrotic tissues instead of across all tissues at once.
• Build the classifier with human and mouse data first, then test whether it still works on a third mammal atlas.
• Replace the ML classifier with a simple gene-set scoring method and compare how much interpretability you gain or lose.

Learn More

• Tabula Sapiens Consortium papers: Search the journal Nature for the main Tabula Sapiens atlas and related methods papers.
• Tabula Muris Consortium papers: Search Nature and Cell for the Tabula Muris single-cell atlas and analysis articles.
• NCBI Gene: Look up orthologs, gene names, and functional summaries at the National Center for Biotechnology Information.
• PubMed: Search review articles on fibroblast heterogeneity, single-cell RNA-seq, and cross-species atlas integration.
• MIT OpenCourseWare: Find free genomics, machine learning, and computational biology course materials for background methods.
• NIH Common Fund Single Cell Analysis Program: Read about single-cell standards, data types, and analysis concepts from NIH resources.