Tumor Margin Spatial Transcriptomics Analysis

ISEF Category: Cellular and Molecular Biology

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

The Hook

Tumors do not grow like smooth balls. Their edges act more like a messy border zone where cells switch behavior fast. That edge can hold clues about invasion, spread, and why some cancers get worse. You can study that border with public data, no wet lab needed.

What Is It?

Spatial transcriptomics measures which genes are active and where they sit in a tissue slice. Think of it like a city map where each block shows the messages cells are sending. Instead of mixing all the cells together, you keep the layout, so you can ask what changes at the tumor edge.

This project focuses on the boundary between tumor and normal tissue. You look for a gene module, which means a group of genes that rise and fall together, in cells near that border. If that module keeps showing up in different tumor types, it may mark cells that help invasion start or spread farther.

Visium and Slide-seq are two ways to make these maps. Visium uses spots on a slide that catch RNA from nearby cells. Slide-seq uses tiny beads with barcodes to track location. Public datasets from these methods let you compare tumor margins without running the experiment yourself.

Why This Is a Good Topic

This is a strong science fair topic because the core question is measurable with public data, and the analysis can be repeated across datasets. You can test whether one boundary-cell gene pattern appears near invasive tumor edges, then see if it holds in more than one cancer type. That connects to real cancer biology, especially how tumors spread. You also learn a real research skill, spatial data analysis, which is useful far beyond one project.

Research Questions

• How does the expression of candidate boundary-cell genes change from the tumor core to the tumor edge?
• What is the effect of tumor subtype on the strength of the boundary-cell gene module?
• Does a boundary-cell score predict known invasive regions better than random gene sets?
• To what extent do Visium and Slide-seq datasets agree on the same edge-associated genes?
• Which genes stay enriched at the tumor margin across multiple cancer types?
• What is the effect of including nearby normal tissue on boundary-cell module detection?
• To what extent does the boundary-cell gene module correlate with histology-based invasion annotations?

Basic Materials

• Computer with at least 16 GB RAM and stable internet access.
• Access to public spatial transcriptomics datasets from GEO, ArrayExpress, or published supplemental files.
• Spreadsheet software for data tracking.
• R or Python installed for analysis.
• Seurat or Scanpy for spatial transcriptomics preprocessing.
• ImageJ for checking tissue images and annotation alignment.
• PubMed access for reading related review articles.

Advanced Materials

• High-memory workstation or cloud computing access for large spatial datasets.
• R with Seurat, SPARK, or Giotto for spatial analysis.
• Python with Scanpy, Squidpy, pandas, and scipy for comparison pipelines.
• Differential expression and gene set enrichment tools.
• Reference bulk RNA-seq or single-cell RNA-seq datasets for cross-validation.
• Histology annotation files, if available from the original studies.
• Public pathway databases such as MSigDB, Reactome, or Gene Ontology.

Software & Tools

• R: Handles spatial transcriptomics workflows, statistics, and plots for comparison across samples.
• Python: Supports data cleaning, module scoring, and cross-dataset analysis.
• Seurat: Processes Visium data, finds spatial patterns, and compares tumor regions.
• Scanpy: Analyzes single-cell and spatial expression matrices with flexible scripting.
• ImageJ: Helps inspect tissue images and line up spatial spots with histology.

Experiment Steps

1. Define the exact biological question, such as whether edge cells share a repeatable invasion-associated gene program across tumor types.
2. Choose public datasets that include clear tissue images, tumor margins, and matching metadata.
3. Decide how you will label core, boundary, and normal regions in each dataset.
4. Build a scoring method for the candidate boundary-cell gene module and decide on comparison controls.
5. Plan cross-dataset validation so one tumor type does not drive the whole result.
6. Set the statistical test you will use to compare edge enrichment, module strength, and invasion annotations.

Common Pitfalls

• Using datasets with weak or missing margin annotations, which makes boundary calls unreliable.
• Mixing Visium and Slide-seq outputs without normalizing for platform differences, which can create fake biological signals.
• Treating low-quality spots as real edge cells, which can inflate the boundary gene score.
• Testing too many gene lists at once, which makes the strongest pattern look better than it is.
• Ignoring tumor subtype or tissue source, which can hide or distort cross-cancer comparisons.

What Makes This Competitive

A competitive version of this project would do more than repeat a known edge signature. You could test whether the same boundary-cell module appears in several cancer types, then check whether it still works after careful normalization and subtype control. Strong entries often use a clear validation plan, compare against random or published gene sets, and report effect sizes, not just p-values. You can also stand out by linking the module to invasion labels, histology features, or a second public dataset.

Project Variations

• Focus on one cancer type, such as breast, colon, or brain tumors, and test whether the boundary-cell module changes with stage.
• Compare Visium and Slide-seq datasets from similar tumor types to see whether platform choice changes the genes you detect at the margin.
• Replace the gene-module score with pathway enrichment analysis to test whether migration, hypoxia, or immune evasion pathways cluster at the edge.

Learn More

• NCBI GEO: Search for public spatial transcriptomics datasets and download expression matrices with sample metadata.
• PubMed: Search for review articles on spatial transcriptomics, tumor microenvironment, and invasion biology.
• NIH NCBI Bookshelf: Find free background chapters on gene expression, cancer biology, and bioinformatics basics.
• Human Protein Atlas: Compare gene expression patterns across normal and cancer tissues.
• MIT OpenCourseWare: Look for free genomics, cancer biology, or computational biology course materials.
• Reactome: Explore curated pathway maps to interpret boundary-cell gene modules.