NOTE: These tutorials are for older versions of Seurat <= v1.2

To access the old version of Seurat, install as follows:

# Enter commands in R (or R studio, if installed)
install.packages("devtools")
library(devtools)
install_github("satijalab/seurat@da6cd08")
library(Seurat) 

If you would like to visualize the Zebrafish embryo (using Jeff Farrell's beautiful RGL functions) in Part 4 of the Zebrafish tutorial, please install an X11 client (i.e. XQuartz) and the rgl package in R using install.packages("rgl").

Getting started with Seurat

The input to Seurat is a normalized gene expression matrix, where the rows are genes, and the columns are single cells. To get started, first install the software, which should take less than a minute if you already have R installed. The tutorials below introduce Seurat through guided analyses of published single cell RNA-seq datasets.

---

Tutorial - Unsupervised clustering and marker discovery

Dataset: 301 single cells across diverse tissues from (Pollen et al., Nature Biotechnology, 2014)
Accompanying data files: input data, R markdown files, pre-computed Seurat objects (place in ~/seurat_files)

First, we introduce the Seurat clustering approach on 301 single cells profiled with the Fluidigm C1

Seurat combines linear and non-linear dimensionality reduction algorithms for unsupervised clustering of single cells. The approach can be described briefly:

1. Identification of highly variable genes
2. Linear dimensional reduction (PCA) on variable genes
3. Determine statistically significant principal components
4. Non-linear dimensional reduction (t-SNE) on the significant PC scores (i.e., spectral t-SNE)
5. Density clustering to classify distinct groups of cells on the t-SNE map
6. Marker discovery, visualization, and downstream analysis

• Optional steps after (2) to grow gene list through PCA projection, and after (4) to project additional cells onto the t-SNE map

We applied this same approach to group ~45,000 single retinal cells, produced by Drop-Seq, into 39 robust clusters. A full tutorial on this dataset is coming soon, and you can explore the technique further at http://www.dropseq.org

---

Tutorial - Spatial Inference of single cell data

Dataset: 851 single cells from Zebrafish embryogenesis (Satija*, Farrell* et al., Nature Biotechnology, 2015)
Accompanying data files: input data, R markdown files, pre-computed Seurat objects (place in ~/seurat_files)

Here, we introduce a computational strategy to infer the original spatial location of a single cell based on its gene expression data. Seurat accomplishes this by leveraging pre-existing in situ hybridization data ('spatial reference map'), which are available or can be easily generated for many developing organisms.