BS6220 Group Assignment: Spatial Transcriptomics Analysis of Hepatocellular Carcinoma Treatment Response

Overview

This project analyzes spatial transcriptomics data from hepatocellular carcinoma (HCC) patients to understand treatment response patterns using 10x Genomics Visium technology. The analysis compares cell type compositions and ligand-receptor interactions across different treatment response categories.

Data Description

• Samples: 26 paired samples from HCC patients (Pre-treatment and Post-treatment)
• Quality Control: 21 samples passed QC (81% pass rate)
• Technology: 10x Genomics Visium spatial transcriptomics
• Normalization: SCTransform (SCT)

Response Categories

Category	Description	Sample Count
NR	Non-Responder	12 samples
Progressor	Disease Progression	2 samples
Super Responder	Excellent Response	8 samples
Unknown	Unclassified	4 samples

Analysis Methods

Cell Type Deconvolution

• RCTD (Robust Cell Type Decomposition)
• cell2location
• CARD (Cell Type Annotation using Reference Dataset)

Ligand-Receptor Interaction Analysis

• CellChat
• CellPhoneDB
• COMMOT (COMMunication analysis by Optimal Transport)

Repository Structure

BS6220/
├── README.md
└── Group Assignment/
    ├── Analysis_Report_Final.md    # Comprehensive analysis report
    ├── README.md
    └── figures/                     # All visualization outputs
        ├── celltype_*.png          # Cell type analysis figures
        ├── lr_*.png                # Ligand-receptor analysis figures
        ├── spatial_*.png           # Spatial visualization figures
        ├── dimplot_*.png           # UMAP clustering figures
        ├── method_comparison_*.png # Tool comparison figures
        └── marker_heatmap.png      # Marker gene heatmap

Key Figures

Spatial Visualizations

• spatial_all_response_types_combined.png - Spatial cluster distribution comparing Pre/Post treatment across response categories
• spatial_clusters_*.png - Individual sample spatial clustering
• spatial_features_*.png - Spatial feature expression patterns
• spatial_colocalization.png - Cell type co-localization analysis

UMAP Clustering

• dimplot_all_response_types_combined.png - UMAP visualization comparing Pre/Post treatment for NR, Progressor, and Super Responder samples

Cell Type Analysis

• celltype_heatmap_response.png - Cell type proportion heatmap by response
• celltype_boxplot_response.png - Cell type distribution boxplots
• celltype_timepoint_interaction.png - Pre/Post treatment comparison

Ligand-Receptor Analysis

• lr_heatmap_response.png - L-R interaction intensity heatmap
• lr_barplot_response.png - L-R interaction frequency
• lr_timepoint_interaction.png - Treatment-induced L-R changes

Method Comparison

• method_comparison_celltype.png - Comparison of cell type deconvolution tools
• method_comparison_lr.png - Comparison of L-R interaction tools

Key Findings

1. Super Responders show distinct spatial clustering patterns with increased immune cell infiltration post-treatment
2. Non-Responders maintain similar cell type compositions before and after treatment
3. Progressors exhibit expansion of tumor-associated cell populations
4. Different deconvolution methods show high concordance for major cell types
5. L-R interaction analysis reveals treatment-induced changes in immune-tumor communication

Technical Notes

• All 26 samples contain SCT (SCTransform) normalized data
• Per-sample analysis approach avoids cross-sample dimension mismatch issues
• Spatial coordinates extracted using GetTissueCoordinates() for visualization

Authors

BS6220 Course Group Assignment

License

This project is for educational purposes as part of the BS6220 course.