Anatomy of a Seurat Object

category_1

category_2

category_3

category_4

Write a description of the lesson here.

Author

Noor Sohail

Published

April 15, 2026

Keywords

keyword_1, keyword_2, keyword_3, keyword_4, keyword_5, keyword_6

Approximate time: XX minutes

Learning objectives

In this lesson, we will:

Learning Objective 1
Learning Objective 2
Learning Objective 3

Overview of lesson

When doing XYZ…

Load dataset

library(Seurat)
library(tidyverse)
crc <- Load10X_Spatial(data.dir = "data/P5CRC_cropped/",
                       bin.size = c(8, 16),
                       slice = "P5CRC")

Anatomy of a Seurat object

As we can see from our Seurat callout, there are a lot of different slots inside our object. Here, we will go through each of the major components of a Seurat object and how you would access key pieces of information.

Assays

Assays are where we can store different counts matrices - we are not forced to keep the same features and variable genes across assays. Each assay will contain it’s own Layers that can be distinct from other assays in the object. This is useful in several different cases:

Multi-modal assays, where you can keep the expression matrices for RNA, ATAC, or protein in a single Seurat object
Storing counts matrices from a variety of different normalization techniques
Batch integration methods will sometimes generate a transformed counts matrix

Here we can print the different assays that exist within our crc object:

Assays(crc)

[1] "Spatial.008um" "Spatial.016um"

We have 2 distinct assays for the different bin sizes, which makes sense because we have different count matrices for our cells based upon the bin size that was selected.

The DefaultAssay() function shows us which assay information will be used in other Seurat function calls, unless explicitly specified otherwise.

DefaultAssay(crc)

[1] "Spatial.008um"

We can also change what our default assay is. Let’s set it to the 016um bins:

DefaultAssay(crc) <- "Spatial.016um"
crc

An object of class Seurat 
36170 features across 97570 samples within 2 assays 
Active assay: Spatial.016um (18085 features, 0 variable features)
 1 layer present: counts
 1 other assay present: Spatial.008um
 2 spatial fields of view present: P5CRC.008um P5CRC.016um

Now we see that the callout says: Active assay: Spatial.016um

`Features` and `Cells`

Our count matrices function as any other matrix does, with rows and columns.

In Seurat, the rows correspond to Features. In the case of spatial transcriptomics, our features are genes. In other experiments, features could refer to chromatin peaks or proteins. The important thing to keep in mind is what technology you are using. Since this is a Visium HD dataset, we are quantifying RNA expression (genes).

We can see what the first few genes/features in our count matrix are:

Features(crc) %>% head()

[1] "SAMD11"  "NOC2L"   "KLHL17"  "PLEKHN1" "PERM1"   "HES4"

As well as see the number of genes that are found in each of our assays:

nrow(crc[["Spatial.008um"]])

[1] 18085

nrow(crc[["Spatial.016um"]])

[1] 18085

The columns correspond to Cells (or samples as it appears in the callout). We can see what the first few cells in our count matrix are:

Cells(crc) %>% head()

[1] "s_016um_00050_00315-1" "s_016um_00064_00214-1" "s_016um_00101_00317-1"
[4] "s_016um_00049_00195-1" "s_016um_00032_00133-1" "s_016um_00061_00268-1"

As well as see the number of cells that are found in each of our assays:

ncol(crc[["Spatial.008um"]])

[1] 77896

ncol(crc[["Spatial.016um"]])

[1] 19674

Exercise 2

What differences do you see between the 8um and 16um bins?

Layers

Layers are our count matrices.

Layers(crc)

[1] "counts"

By default, Seurat uses the following naming convention for the counts matrices within an Assay:

Table 1: Description of each Layer() count matrix

Layer	Description
counts	Raw counts
data	Normalized counts
scale.data	Scaled‑normalized counts

You may notice that our Seurat object only contains counts right now. This is because we have not run any normalization steps yet (we will discuss how to do so in future lessons).

Using the LayerData() function we can access the entire counts matrix. Furthermore, we can specify the assay if we would prefer to not use the DefaultAssay.

LayerData(crc, 
          assay = "Spatial.016um", 
          layer = "count")[1:5, 1:5]

Table 2: First 5 rows and columns of raw counts matrix

	s_016um_00050_00315-1	s_016um_00064_00214-1	s_016um_00101_00317-1	s_016um_00049_00195-1	s_016um_00032_00133-1
SAMD11	0	0	0	0	0
NOC2L	0	0	0	0	0
KLHL17	0	0	0	0	0
PLEKHN1	0	0	0	0	0
PERM1	0	0	0	0	0

By printing the first 5 features and cells in our object (for easier visualization). We can see that we are working with whole numbers which reinforces the idea that this is the raw data, with no transformations having been applied.

Spatial fields

We do not just have expression data associated with our sample, we also have the spatial slide that comes with its own set of values and information.

For example we can grab the x,y coordinates of each bin using the GetTissueCoordinates() function.

GetTissueCoordinates(crc) %>% View()

Table 3: Coordinates on spatial slide for each cell

	x	y	cell
s_016um_00050_00315-1	62715.76	61103.62	s_016um_00050_00315-1
s_016um_00064_00214-1	56816.46	60261.12	s_016um_00064_00214-1
s_016um_00101_00317-1	62844.88	58123.60	s_016um_00101_00317-1
s_016um_00049_00195-1	55702.46	61133.16	s_016um_00049_00195-1
s_016um_00032_00133-1	52074.99	62111.72	s_016um_00032_00133-1

Or visualize what our slide looks like with SpatialDimPlot():

SpatialDimPlot(crc, pt.size.factor = 15)

Figure 1: Spatial visualization of spots with `SpatialDimPlot()`

Metadata

Seurat automatically creates some metadata for each of the cells when the object is created. This information is stored in the @meta.data slot within the Seurat object. The rownames are automatically set to be the cell names.

crc@meta.data %>% View()

Table 4: Seurat default @meta.data

	orig.ident	nCount_Spatial.008um	nFeature_Spatial.008um	nCount_Spatial.016um	nFeature_Spatial.016um
s_008um_00078_00444-1	s	65	57	NA	NA
s_008um_00128_00278-1	s	1300	906	NA	NA
s_008um_00052_00559-1	s	128	121	NA	NA
s_008um_00121_00413-1	s	538	326	NA	NA
s_008um_00167_00326-1	s	44	39	NA	NA

What does each column represent?

Table 5: Columns automatically populated in @meta.data

Column	Description
orig.ident	Sample identity if known; defaults to “s”
nCount_RNA	Number of UMIs per cell
nFeature_RNA	Number of genes detected per cell

While it may seem intimidating at first, the important thing to remember is that this is a dataframe. Therefore can modify and work with this dataframe just like we would any other in R! For example, we can set our orig.ident column to be our sample name rather than “s”.

crc@meta.data$orig.ident <- "P5CRC"

crc@meta.data %>% View()

Table 6: Seurat default @meta.data after updating orig.ident

	orig.ident	nCount_Spatial.008um	nFeature_Spatial.008um	nCount_Spatial.016um	nFeature_Spatial.016um
s_008um_00078_00444-1	P5CRC	65	57	NA	NA
s_008um_00128_00278-1	P5CRC	1300	906	NA	NA
s_008um_00052_00559-1	P5CRC	128	121	NA	NA
s_008um_00121_00413-1	P5CRC	538	326	NA	NA
s_008um_00167_00326-1	P5CRC	44	39	NA	NA

Additionally, we do not have use the @meta.data each time we want to access a single column. We can use the $ follow by the column name as a shorthand.

crc$nCount_Spatial.008um %>% head()

s_008um_00078_00444-1 s_008um_00128_00278-1 s_008um_00052_00559-1 
                   65                  1300                   128 
s_008um_00121_00413-1 s_008um_00167_00326-1 s_008um_00202_00633-1 
                  538                    44                   365

Idents

The cell identities are stored as Idents(), which contain the default way to label cells. For example, if we wanted to label each cell by which sample they came from we could run:

Idents(crc) <- "orig.ident"
Idents(crc) %>% head()

s_008um_00078_00444-1 s_008um_00128_00278-1 s_008um_00052_00559-1 
                P5CRC                 P5CRC                 P5CRC 
s_008um_00121_00413-1 s_008um_00167_00326-1 s_008um_00202_00633-1 
                P5CRC                 P5CRC                 P5CRC 
Levels: P5CRC

Where we see that the identities of the cells are the value stored in the @meta.data column orig.ident.

Back to Lesson >>

Back to Schedule

Reuse

CC-BY-4.0

--- title: "Anatomy of a Seurat Object" description: | Write a description of the lesson here. author: - Noor Sohail date: "2026-04-15" categories: - category_1 - category_2 - category_3 - category_4 keywords: - keyword_1 - keyword_2 - keyword_3 - keyword_4 - keyword_5 - keyword_6 license: "CC-BY-4.0" editor_options: markdown: wrap: 72 --- ```{r} #| label: load_libraries_data #| echo: false # Load libraries and data ``` Approximate time: XX minutes ## Learning objectives In this lesson, we will: - Learning Objective 1 - Learning Objective 2 - Learning Objective 3 ## Overview of lesson When doing XYZ... ## Load dataset ```{r} #| label: crc_Load10X_Spatial library(Seurat) library(tidyverse) crc <- Load10X_Spatial(data.dir = "data/P5CRC_cropped/", bin.size = c(8, 16), slice = "P5CRC") ``` ## Anatomy of a Seurat object As we can see from our Seurat callout, there are a lot of different slots inside our object. Here, we will go through each of the major components of a Seurat object and how you would access key pieces of information. ### Assays `Assays` are where we can store different counts matrices - we are not forced to keep the same features and variable genes across assays. Each assay will contain it's own `Layers` that can be distinct from other assays in the object. This is useful in several different cases: - Multi-modal assays, where you can keep the expression matrices for RNA, ATAC, or protein in a single Seurat object - Storing counts matrices from a variety of different normalization techniques - Batch integration methods will sometimes generate a transformed counts matrix Here we can print the different assays that exist within our `crc` object: ```{r} #| label: crc_assays Assays(crc) ``` We have 2 distinct assays for the different bin sizes, which makes sense because we have different count matrices for our cells based upon the bin size that was selected. The `DefaultAssay()` function shows us which assay information will be used in other Seurat function calls, unless explicitly specified otherwise. ```{r} #| label: crc_DefaultAssay DefaultAssay(crc) ``` We can also change what our default assay is. Let's set it to the `016um` bins: ```{r} #| label: crc_DefaultAssay_callout DefaultAssay(crc) <- "Spatial.016um" crc ``` Now we see that the callout says: `Active assay: Spatial.016um ` ### `Features` and `Cells` Our count matrices function as any other matrix does, with rows and columns. In Seurat, the **rows** correspond to `Features`. In the case of spatial transcriptomics, our features are genes. In other experiments, features could refer to chromatin peaks or proteins. The important thing to keep in mind is what technology you are using. Since this is a Visium HD dataset, we are quantifying RNA expression (genes). We can see what the first few genes/features in our count matrix are: ```{r} #| label: crc_features Features(crc) %>% head() ``` As well as see the number of genes that are found in each of our assays: ```{r} #| label: crc_nrow nrow(crc[["Spatial.008um"]]) nrow(crc[["Spatial.016um"]]) ``` The **columns** correspond to `Cells` (or samples as it appears in the callout). We can see what the first few cells in our count matrix are: ```{r} #| label: crc_cells Cells(crc) %>% head() ``` As well as see the number of cells that are found in each of our assays: ```{r} #| label: crc_ncol ncol(crc[["Spatial.008um"]]) ncol(crc[["Spatial.016um"]]) ``` ::: callout-tip # [**Exercise 2**](03_loading_spatial_data-Answer_key.qmd#exercise-2) 3. What differences do you see between the `8um` and `16um` bins? ::: ### Layers Layers are our count matrices. ```{r} #| label: crc_layers Layers(crc) ``` By default, Seurat uses the following naming convention for the counts matrices within an `Assay`: | Layer | Description | |-------------|---------------------------------| | counts | Raw counts | | data | Normalized counts | | scale.data | Scaled‑normalized counts | : Description of each `Layer()` count matrix {#tbl-layer_descriptions} You may notice that our Seurat object only contains `counts` right now. This is because we have not run any normalization steps yet (we will discuss how to do so in future lessons). Using the `LayerData()` function we can access the entire counts matrix. Furthermore, we can specify the `assay` if we would prefer to not use the `DefaultAssay`. ```{r} #| label: crc_LayerData #| eval: false LayerData(crc, assay = "Spatial.016um", layer = "count")[1:5, 1:5] ``` ```{r} #| label: tbl-crc_LayerData_kable #| tbl-cap: First 5 rows and columns of raw counts matrix #| echo: false LayerData(crc, assay = "Spatial.016um", layer = "count")[1:5, 1:5] %>% as.data.frame() %>% knitr::kable() ``` By printing the first 5 features and cells in our object (for easier visualization). We can see that we are working with **whole numbers** which reinforces the idea that this is the **raw data**, with no transformations having been applied. ### Spatial fields We do not just have expression data associated with our sample, we also have the spatial slide that comes with its own set of values and information. For example we can grab the x,y coordinates of each bin using the `GetTissueCoordinates()` function. ```{r} #| label: crc_GetTissueCoordinates #| eval: false GetTissueCoordinates(crc) %>% View() ``` ```{r} #| label: tbl-crc_GetTissueCoordinates_dt #| tbl-cap: Coordinates on spatial slide for each cell #| echo: false GetTissueCoordinates(crc) %>% head(5) %>% knitr::kable() ``` Or visualize what our slide looks like with `SpatialDimPlot()`: ```{r} #| label: fig-crc_SpatialDimPlot #| fig-cap: Spatial visualization of spots with `SpatialDimPlot()` SpatialDimPlot(crc, pt.size.factor = 15) ``` ### Metadata **Seurat automatically creates some metadata** for each of the cells when the object is created. This information is stored in the `@meta.data` slot within the Seurat object. The rownames are automatically set to be the cell names. ```{r} #| label: crc_meta_data_1 #| eval: false crc@meta.data %>% View() ``` ```{r} #| label: tbl-crc_meta_data_1_dt #| tbl-cap: Seurat default `@meta.data` #| echo: false crc@meta.data %>% head(5) %>% knitr::kable() ``` What does each column represent? | Column | Description | |---------------|-----------------------------------------------| | orig.ident | Sample identity if known; defaults to “s” | | nCount_RNA | Number of UMIs per cell | | nFeature_RNA | Number of genes detected per cell | : Columns automatically populated in `@meta.data` {#tbl-metadata_cols} While it may seem intimidating at first, the important thing to remember is that this is a dataframe. Therefore can modify and work with this dataframe just like we would any other in R! For example, we can set our `orig.ident` column to be our sample name rather than "s". ```{r} #| label: crc_meta_data_orig_ident crc@meta.data$orig.ident <- "P5CRC" ``` ```{r} #| label: crc_meta_data_2 #| eval: false crc@meta.data %>% View() ``` ```{r} #| label: tbl-crc_meta_data_2_dt #| tbl-cap: Seurat default `@meta.data` after updating `orig.ident` #| echo: false crc@meta.data %>% head(5) %>% knitr::kable() ``` Additionally, we do not have use the `@meta.data` each time we want to access a single column. We can use the `$` follow by the column name as a shorthand. ```{r} #| label: crc_dollar_notation crc$nCount_Spatial.008um %>% head() ``` ### Idents The cell identities are stored as `Idents()`, which contain the default way to label cells. For example, if we wanted to label each cell by which sample they came from we could run: ```{r} #| label: crc_idents Idents(crc) <- "orig.ident" Idents(crc) %>% head() ``` Where we see that the identities of the cells are the value stored in the `@meta.data` column `orig.ident`. *** [Back to Lesson >>](03_loading_spatial_data.qmd) [Back to Schedule](../schedule/schedule.qmd)

Learning objectives

Overview of lesson

Load dataset

Anatomy of a Seurat object

Assays

Features and Cells

Layers

Spatial fields

Metadata

Idents

Reuse

`Features` and `Cells`