This project explores transcriptomic data from blood and brain samples to examine cell-type-specific gene expression patterns relevant to inflammation and neurodegeneration. The work focuses on exploratory analysis, clustering, and visualization to understand how molecular profiles can inform precision and preventive medicine research.
- GSE63060 - Normalized blood transcriptomic data
- GSM5645890 - Brain metastasis single-cell RNA-seq data with cell type annotations
- Visualize and cluster cell types based on gene expression
- Compare inflammatory and Alzheimer’s-related genes across cell types
- Use PCA and KMeans to define digital phenotypes
- Score inflammatory burden per cell type
- Allow interactive gene panel selection
| Method | Purpose |
|---|---|
| Data normalization | Ensures comparability across samples |
| Gene symbol mapping | Maps probe IDs to gene names |
| Heatmaps & PCA | Highlights top 20 variable genes |
| Alzheimer’s panel | Focuses on genes like APP, MAPT, APOE, TREM2 |
| Inflammation score | Combines IL1B, IL6, TNF, CXCL8 expression |
| Clustering | KMeans on PCA-transformed data |
Here are a few key visuals generated from the analysis:
-
Alzheimer’s Genes Across Cell Types
-
Inflammation Scores by Cell Type
-
Top Variable Genes Heatmap
-
Cell Type Clusters (PCA)
These figures help highlight which cell types may be more active in inflammation or neurodegenerative pathways, offering insights into disease mechanisms from a systems biology view.
- Open the notebook:
Blood_Transcriptomics_Analysis.ipynb - Follow the markdown explanations step-by-step
- Choose your own genes of interest when prompted
- Review the plots and summary results
This project reflects principles of data-driven systems medicine:
- Integrates data from bulk and single-cell studies
- Uses dimension reduction + unsupervised clustering
- Builds quantitative cell-type signatures for immune/inflammatory profiling
It demonstrates how large-scale data can inform precision diagnostics and cell-specific insights — bridging data science and translational research.
- Add more patient data over time to track changes (like in aging or treatment)
- Connect gene expression to real-world health info (e.g., symptoms, diagnosis)
- Build a basic version of a digital twin – a personal model using gene activity
- Try more clustering or machine learning methods for deeper patterns
- Explore other diseases beyond inflammation or Alzheimer’s
Zareen Rahman
Motivated by systems biology, and currently exploring opportunities to assist with translational research.
MIT License – feel free to use or extend this work.