Spatial Transcriptomics Analysis

This project focuses on the analysis of spatial transcriptomics data using a 10X Visium dataset of a mouse brain, utilizing the Seurat package in R.

Key Tasks:

1. Spatial Transcriptomics:

   • Properties of the slides were analyzed, including the size of the point, distance between points, and the number of points used in the technology.
   • The resolution of the spatial transcriptomics technology was compared to the size of an average eukaryotic cell, considering its implications for data analysis.
   • An initial overview of the data was performed, including visualization of the sample image, coordinates of the spots, and one gene expression matrix.

2. Spatial Transcriptomics Data in Seurat:

   • The data was loaded using Seurat's Load10X_Spatial function.
   • The Seurat object was inspected to identify where the gene expression data and tissue image are stored.
   • Gene expression for two randomly chosen genes was visualized in the tissue.

3. Data Preprocessing:

   • The data was filtered with reasonable cut-off values, justified with appropriate plots. The filtering thresholds were compared to those used in scRNA-seq from Project 1.
   • SCTransform was used for preprocessing, replacing some steps from Project 1.

4. Dimensionality Reduction, Clustering, and Visualization:

   • Dimensionality reduction was performed using PCA and UMAP, and a plot was used to explain the choice of dimensions.
   • Clustering was done based on PCA results and visualized in both 2D UMAP space and on the tissue slide.

5. Differential Expression Analysis:

   • Differentially expressed genes (DEGs) were analyzed based on clustering and spatial patterning.
   • The top three spatially variable features were identified and visualized on the tissue slide. Comparisons were made with the DEG results from clustering.

6. Merging the Data:

   • Two datasets were merged without batch correction, followed by pre-processing, dimensionality reduction, and clustering.
   • Merging with batch correction was performed, and the results were displayed in UMAP space.
   • Batch effects were assessed and resolved, if necessary, for further processing.

7. Cell-type Identification:

   • Automatic annotation was performed using Data Integration with scRNA-seq data, and the labels were transferred to the spatial transcriptomics dataset. The annotation was visualized in UMAP space.
   • Manual annotation was performed by identifying marker genes for selected cell types, and gene expression for these markers was visualized on UMAP plots and tissue slides.

8. Deconvolution:

   • Deconvolution was performed using the SCDC package. The methodology behind deconvolution was explained, including its necessity and limitations.
   • A reference dataset was prepared, and genes for deconvolution were selected based on DEG analysis.
   • An ExpressionSet object was created for the reference and query data, and deconvolution was performed. The distribution of cell types on the tissue slide was visualized and compared to previous results.

R Script:

The R script used for analysis is named Spatial-Transcriptomics-script.

Results:

The resulting plots and images can be found in the png/ directory of this repository.

References:

For more details, refer to the official Seurat documentation for Spatial Transcriptomics: Seurat Spatial Vignette.