Expression Analysis from Matrix
The workflow Expression Analysis from Matrix takes one or more Expression Matrices (
) as input and performs quality control, normalization, clustering and cell type prediction. The workflow uses iterate functionality and allows for a combined analysis of multiple samples to produce:
- a single, multi-sample, normalized Expression Matrix ();
- a Dimensionality Reduction Plot (
) associated with the automated clusters, predicted cell types and additional cell annotations;
- a Heat Map (
), a Dot Plot (
), and a Violin Plot (
) with the predicted cell types as cell groups;
- a Cell Abundance Heat Map () with the automated clusters and predicted cell types as cell groups.
The workflow can be found in the toolbox here:
Workflows (
) | From Imported Data (
) | Expression Analysis from Matrix (
)
If you are connected to a CLC Server via your Workbench, you will be asked where you would like to run the analysis. We recommend that you run the analysis on a CLC Server when possible.
Choose either one or more Expression Matrix () or Select files for import and select the expression matrix format that is compatible with the selected input. Read more about the import options in Data import.
The workflow offers a number of options described below. Note that not all parameters can be configured. Open parameters indicate places where customization may be necessary for different samples, but default settings are suitable in most cases.
The workflow can be run using Single Cell hg38 (Ensembl) or Single Cell Mouse (Ensembl) reference data sets (see The Reference Data Manager).
|
Note: Reference data elements cannot be configured during workflow execution. If other elements than those provided in the default reference data sets are needed a custom reference data set can be used, see
https://resources.qiagenbioinformatics.com/manuals/clcgenomicsworkbench/current/index.php?manual=Custom_Sets.html. When creating custom reference data sets, the chosen gene track needs to match the gene annotations used for training the provided Cell Type Classifier ( |
), see
The workflow allows the analysis of multiple samples and you can specify metadata during workflow execution. This is converted to cell annotations and can be used for coloring the cells in the Dimensionality Reduction Plot. However, the workflow expects each sample to be present in just one Expression Matrix, and attempting to define batch units containing more than one Expression Matrix will lead to a failure during execution.
For more details on configuring workflow execution with metadata, see https://resources.qiagenbioinformatics.com/manuals/clcgenomicsworkbench/current/index.php?manual=Batching_part_workflow.html. Make sure to inspect the batch overview to check that the analysis will be performed correctly.
For quality control a number of options exist. The option to remove empty droplets is not suitable for protocols that do not use droplets, and removing barcodes with low number of reads or expressed features might be more appropriate. Quality Control (QC) uses the number of reads mapped to the mitochondria, and for this the name of the mitochondria chromosome needs to be provided. The default value is often the correct name. After quality control, the matrices are collected and normalized jointly. Note that batch correction is not performed. Read more about QC and normalization in Cell preparation.
For clustering and creation of the Dimensionality Reduction Plot plot, it is possible to restrict analysis to highly variable genes. The data is then projected to a lower dimensional space using PCA. You can read about this feature in Feature selection and PCA.
The expression plots (Heat Map and Dot Plot) and Cell Abundance Heat Map group the cells based on the high confidence predicted cell types ("Cell type (high confidence)"). The Cell Abundance Heat Map additionally groups the cells based on the automated clusters obtained with resolution 1.0 ("Leiden (resolution=1.0)"). Any of these groups can be changed to:
- all predicted cell types ("Cell type (all)");
- automated clusters obtained with a different resolution
("Leiden (resolution=)"). All resolutions
are produced, in steps of 
.
Subsections