Thesis Project: Spatial Transcriptomics Analysis

Short Specialization Degree in Omics Data Analysis, University of Padua

This repository contains the code and methods used in my thesis project for the short specialization degree in Omics Data Analysis at the University of Padua. The project focuses on spatial transcriptomics, utilizing publicly available datasets from the spatialLIBD package.

Aim of the project: spatial transcriptomics analysis of Human DLPFC

This project uses spatial transcriptomics data from the Human Dorsolateral Prefrontal Cortex (DLPFC) to analyze gene expression patterns in specific cortical layers and identify spatially variable genes (SVGs). Particularly, the differences between the "second" layer of the DLPFC and the white matter are analyzed as a proof-of-principle of the methodology. A pdf presentation is also included in this repository, explaining the various steps of the analyses, outputs (tables and plots), and the conclusions.

Requirements

This analysis relies on the following R packages:

• spatialLIBD - Data handling and spatial visualization.
• SingleCellExperiment - Handling single-cell data structures.
• SpatialExperiment - Defines an S4 class for storing data from spatial -omics experiments.
• nnSVG - Detection of spatially variable genes.
• PlackettLuce - Ranking and preference analysis.
• dplyr and ggplot2 - Data manipulation and visualization.
• rmarkdown and knitr - Document generation and table rendering.

Analysis Workflow

1. Data Loading and Preparation: the Visium spatial transcriptomics data for DLPFC samples is loaded and updated using functions from the spatialLIBD package. This step includes converting single-cell data objects into a format suitable for spatial analysis.
2. Spatial Data Exploration and Visualization: this section enables a deeper understanding of the spatial organization of gene expression across different regions of the DLPFC, aiding in layer-specific analysis and interpretation

• Spatial Layer Mapping: Visualizes cellular distribution across cortical layers for selected samples, using color-coded clusters to represent distinct layers or regions.
• Marker Gene Expression: Examines spatial expression patterns of specific marker genes (e.g., white matter markers) to identify regions of high enrichment, helping to reveal functional areas within the tissue.
• Layer-Specific Enrichment: Quantifies gene enrichment across layers, providing insights into the predominant layer for each marker gene through plots and statistical summaries.

4. Spatial Registration: single-nucleus RNA-seq data is loaded and used to perform spatial registration, aligning gene expression profiles from single cells to the spatial transcriptomics data. This process computes pseudo-bulking, Bayesian correlation, and t-statistics to enhance spatial resolution.
5. Spatially Variable Gene Analysis: the nnSVG package is used to identify spatially variable genes (SVGs) within specific cortical layers, including white matter and Layer 2. The script demonstrates filtering and downsampling steps to ensure robust SVG analysis and addresses potential issues with duplicate coordinates.
6. Ranking and Comparison of Genes: the PlackettLuce package is used to rank genes based on expression in white matter and Layer 2, generating coefficients for each gene to assess its relative importance. Plots display these rankings for visual comparison between the two layers.

Example Visualizations

This script includes visualization functions to:

• Display gene expression patterns for specific markers across spatial coordinates.
• Plot layer-specific gene enrichment and SVG patterns for comparison.

Output

The analysis generates:

• Tables: Summarizing significant genes, spatial registration results, and annotated cell types.
• Plots: Gene expression distribution across layers and ranked gene importance for white matter and Layer 2.