Skip to contents

Get started

Last compiled: 26 September 2023

getting_started.rmd

semla is a toolbox for processing, analysis and visualization of spatially resolved transcriptomics data.

Installation

To install semla, run the following command from R:

remotes::install_github("ludvigla/semla")

Load libraries

Load data

Currently, semla exclusively supports 10x Visium data. The easiest way to load data is to use the ReadVisiumData() function. First, you need a couple of files output by the 10x Genomics Space Ranger command line tool. These files are then assembled into a data.frame-like object with one sample per row that can be used by ReadVisiumData().

  • samples : file paths to expression matrices

  • imgs : file paths to H&E images

  • spotfiles : file paths to spot coordinate tables

  • json : file paths to scale factor json files

We can also add any number of columns to our infoTable which will be passed as meta data columns in our Seurat object. This can for example be useful when you want to add clinical metadata, experimental information etc. The example data below is from two Visium data sets, one mouse brain and one mouse colon data set. Here we add a sample_id column describing what tissue type we have.

data_root_directory <- file.path(system.file("extdata", package = "semla"), "*")

samples <- Sys.glob(paths = file.path(data_root_directory, 
                                      "filtered_feature_bc_matrix.h5"))

imgs <- Sys.glob(paths = file.path(data_root_directory, 
                                   "spatial", "tissue_lowres_image.jpg"))

spotfiles <- Sys.glob(paths = file.path(data_root_directory, 
                                        "spatial", "tissue_positions_list.csv"))

json <- Sys.glob(paths = file.path(data_root_directory, 
                                   "spatial", "scalefactors_json.json"))
infoTable <- tibble(samples, imgs, spotfiles, json, # Add required columns
                    sample_id = c("mousebrain", "mousecolon")) # Add additional column


Now we are ready to load the 10x Visium data into a Seurat object:

se <- ReadVisiumData(infoTable)
se
## An object of class Seurat 
## 188 features across 5164 samples within 1 assay 
## Active assay: Spatial (188 features, 0 variable features)


You can use most Seurat methods as long as they do not require spatial information. It’s important to know that the Seurat R package stores spatial information (such as H&E images and coordinates) differently than semla, meaning that Seurat functions such as SpatialFeaturePlot() and FindSpatiallyVariableFeatures() will not work. semla comes with it’s own suite of tools and methods that you can learn more about in other tutorials.


If you need to access the spatial data, such as images and coordinates, you can use the GetStaffli function on your Seurat object.

spatial_data <- GetStaffli(se)


The H&E images are only loaded when LoadImages is called on the Seurat object.

se <- LoadImages(se)
ImagePlot(se)

This should be enough to get you started with the rest of the tutorials!

NB: An important thing to know is that some generic functions, such as subset and merge should not be used with semla. Instead, use the SubsetSTData() and MergeSTData() functions - See the vignette Subset/merge.

Make Seurat object compatible with semla

If you have a Seurat object with Visium data that was prepared using Read10X_Image, it is possible to add a Staffli object for compatibility with semla using the UpdateSeuratForSemla function.

By doing so, you can now use both spatial visualization functions from Seurat and semla.

## ── Installed datasets ───────────────────────────────────── SeuratData v0.2.2 ──
## ✔ stxBrain 0.1.1
## ────────────────────────────────────── Key ─────────────────────────────────────
## ✔ Dataset loaded successfully
## ❯ Dataset built with a newer version of Seurat than installed
## ❓ Unknown version of Seurat installed
InstallData("stxBrain")
# Load example Seurat object
brain <- LoadData("stxBrain", type = "anterior1")

# Make Seurat object compatible with semla
brain_semla <- UpdateSeuratForSemla(brain)
# Load H&E images
brain_semla <- LoadImages(brain_semla)

# Plot with semla
MapFeatures(brain_semla, features = "nFeature_Spatial", 
            image_use = "raw", override_plot_dims = TRUE) & ThemeLegendRight()

# Plot with Seurat
SpatialFeaturePlot(brain_semla, features = "nFeature_Spatial")