04-08-singleR.Rmd

# (PART) Futher Analysis {-} 


# SingleR {#singler}

```{r results='hide', message=FALSE, warning=FALSE}
#install.packages("BiocManager")
#BiocManager::install(c("SingleCellExperiment","SingleR","celldex"),ask=F)
library(SingleCellExperiment)
library(SingleR)
library(celldex)
```

In this workshop we have focused on the Seurat package.  However, there is another whole ecosystem of R packages for single cell analysis within Bioconductor.  We won't go into any detail on these packages in this workshop, but there is good material describing the object type online : [OSCA](https://robertamezquita.github.io/orchestratingSingleCellAnalysis/data-infrastructure.html).

For now, we'll just convert our Seurat object into an object called SingleCellExperiment.  Some popular packages from Bioconductor that work with this type are Slingshot, Scran, Scater.

```{r}
sce <- as.SingleCellExperiment(seurat_object)
sce
```

We will now use a package called SingleR to label each cell.  SingleR uses a reference data set of cell types with expression data to infer the best label for each cell.  A convenient collection of cell type reference is in the `celldex` package which currently contains the follow sets:

```{r}
ls('package:celldex')
```
In this example, we'll use the `HumanPrimaryCellAtlasData` set, which contains high-level, and fine-grained label types. Lets download the reference dataset

```{r}
# This too is a sce object,
# colData is equivalent to seurat's metadata
ref.set <- celldex::HumanPrimaryCellAtlasData()
```

The "main" labels.
```{r}
unique(ref.set$label.main)
```

An example of the types of "fine" labels.

```{r}
head(unique(ref.set$label.fine))
```

Now we'll label our cells using the SingleCellExperiment object, with the above reference set.

```{r}
pred.cnts <- SingleR::SingleR(test = sce, ref = ref.set, labels = ref.set$label.main)
```

Keep any types that have more than 10 cells to the label, and put those labels back on our Seurat object and plot our on our umap.

<!-- previously this used first.lables, which no longer seems to exist in pred.cnts -->
<!-- labels and pruned labels are identical in this data --> 

```{r}
lbls.keep <- table(pred.cnts$labels)>10
seurat_object$SingleR.labels <- ifelse(lbls.keep[pred.cnts$labels], pred.cnts$labels, 'Other')
DimPlot(seurat_object, reduction='umap_harmony', group.by='SingleR.labels')
```

Compare cell labels by different annotation methods:

```{r}
DimPlot(seurat_object,group.by = "RNA_snn_res.0.5",reduction = "umap_harmony")

DimPlot(seurat_object,group.by = "cell",reduction = "umap_harmony")



```