Script2_Overlay_scRNAseq_onto_Visium.Rmd

---
title: "Faiz_Strokedata_Part2_Overlay_scRNAseq_on_Visium"
author: "NS"
date: '2022-03-10'
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

## Load Libraries:
```{r, echo=FALSE}

library(Seurat)
library(tidyverse)
library(ggplot2)
library(patchwork)
library(dplyr)
library(hdf5r)
```

## Import the day10_Visium_data:
```{r setup, include=FALSE}

#Define the path to the visium data directory
data_dir_d10 <- "/path/to/data/directory"

#List the files 
list.files(data_dir_d10, all.files = TRUE)

#Read the spatial data using Seurat
FV.d10 <- Load10X_Spatial(data.dir = data_dir_d10)

#Normalize the data (sctransform normalizes the data, detects high-variance features, and stores the data in the SCT assay)
FV.d10 <- SCTransform(FV.d10, assay = "Spatial", verbose = FALSE)

#Dimensionality reduction, clustering, and visualization
FV.d10<- RunPCA(FV.d10, assay = "SCT", verbose = FALSE)
FV.d10<- FindNeighbors(FV.d10, reduction = "pca", dims = 1:30)
FV.d10<- FindClusters(FV.d10, verbose = FALSE)
FV.d10<- RunUMAP(FV.d10, reduction = "pca", dims = 1:30)

#Visualize the results of the clustering 
P1.d10 <- DimPlot(FV.d10, reduction = "umap", label = TRUE)
P2.d10 <- SpatialDimPlot(FV.d10, label = TRUE, label.size = 3)
P1.d10 + P2.d10
```

## Integration Visium data with the scRNA seqdata:
```{r setup, include=FALSE}

#1) Define the reference and query datasets
query1 = FV.d10
ref1 = seu.combined

#2) Pre-process the scRNA-seq reference (note that setting ncells=3000 normalizes the full dataset but learns noise models on 3k cells, this speeds up SCTransform dramatically with no loss in performance)

ref1 <- SCTransform(ref1, ncells = 3000, verbose = FALSE) %>%
  RunPCA(verbose = FALSE) %>%
  RunUMAP(dims = 1:10)

#3) Find transfer anchors
anchors.1 <- FindTransferAnchors(reference = ref1, query = query1, normalization.method = "SCT")

#4) Transfer the identified labels/anchors (The procedure outputs, for each spot, a probabilistic classification for each of the scRNA-seq derived classes)
predictions.1 <- TransferData(anchorset = anchors.1, refdata = ref1$seurat_clusters, prediction.assay = TRUE, weight.reduction = query1[["pca"]], dims = 1:30) # note that you need to use 'refdata = ref1$new_clusters' for plotting with the corrected cluster.id. Please check Script 3 for more information. 
                              

#5) We add these predictions as a new assay in the Seurat object.
query1[["predictions"]] <- predictions.1

#Now we get prediction scores for each spot for each class. 
DefaultAssay(query1) <- "predictions"


#6) Visualize (see how scRNA-seq driven cluster map onto the Visium dataset)
## this code is also presented in script 3, as 'Figure 2-part F'
                                             
SpatialFeaturePlot(query1,features = c( "1","2","3","4","5","6","7", "8") , # for cluster.no starting from 1 please check script-3 
                   pt.size.factor = 1.4, ncol = 4, crop = TRUE)

```