• Product manuals

Browse the manual

• Introduction
  • The concept of CLC Single Cell Analysis Module
  • Contact information
  • System requirements
  • Installing modules
    • Licensing modules
    • Uninstalling modules
  • Installing server extensions
    • Licensing server extensions
• Reference data management
  • QIAGEN Sets
• Running tools and workflows
• Data import
  • Import Immune Reference Segments
    • IMSEQ
    • IMGT
    • Output from Import Immune Reference Segments
    • Filtering gene segments
  • Import Cell Annotations
  • Import Cell Clusters
  • Import Cell Clonotypes
  • Import Expression Matrix
    • HDF5 formats
    • Other formats
  • Import Peak Count Matrix
  • Import Space Ranger
  • Cell format in importers
  • On-the-fly import in workflows
• Data export
  • Export Cell Ranger HDF5 Expression Matrix
  • Export AnnData Expression Matrix
  • Export h5Seurat Expression Matrix
  • Export Loom Expression Matrix
  • Export MEX Expression Matrix
  • Export Plain Text Table Expression Matrix
  • Export Peak Count Matrix
• Prepare Reads
  • Annotate Single Cell Reads
    • Read structure
    • Barcodes from names
    • Barcode correction
    • Sample name
    • The output of Annotate Single Cell Reads
• Creating a Gene Expression Matrix
  • Single Cell RNA-Seq Analysis
    • The Single Cell RNA-Seq Analysis report
    • The Single Cell RNA-Seq Analysis algorithm
  • QC for Single Cell
    • Empty droplets filter
    • Count-based and extra-chromosomal filters
    • Doublets filter
    • The output of QC for Single Cell
    • Choosing barcodes to retain
    • Cell calling
    • Automatic thresholds
    • Doublet calling
  • Demultiplex Parse Bio Samples
  • Normalize Single Cell Data
    • When is batch correction appropriate?
    • The output of Normalize Single Cell Data
    • The Normalize Single Cell Data algorithm
  • The Expression Matrix element
• Cell Type Classification
  • Browse QIAGEN Cell Ontology
  • Predict Cell Types
    • The output of Predict Cell Types
    • Cell type refinement
  • Train Cell Type Classifier
    • Features used for training and prediction
    • The output of Train Cell Type Classifier
    • SVMs for cell type classification
  • Update Cell Type Classifier
  • The Cell Type Classifier element
• Expression Analysis
  • Differential Expression for Single Cell
    • The output of Differential Expression for Single Cell
    • The differential expression algorithm
  • Create Expression Plot
    • The Heat Map output of Create Expression Plot
    • The Dot Plot output of Create Expression Plot
    • The Violin Plot output of Create Expression Plot
• Velocity Analysis
  • Single Cell Velocity Analysis
    • The output of Single Cell Velocity Analysis
    • The velocity estimation algorithm
  • Differential Velocity for Single Cell
  • Score Velocity Genes
    • The output of Score Velocity Genes
  • Create Phase Portrait Plot
  • The Velocity Matrix element
• Spatial Transcriptomics
  • The Spatial Transcriptomics Plot element
• Chromatin Accessibility
  • Single Cell ATAC-Seq Analysis
    • The output of Single Cell ATAC-Seq Analysis
    • The report output from Single Cell ATAC-Seq Analysis
    • The Single Cell ATAC-Seq Analysis algorithm
  • Differential Accessibility for Single Cell
    • The differential accessibility algorithm
  • Create Peak Graph Track
  • The Peak Count Matrix element
• Immune Repertoire
  • Single Cell V(D)J-Seq Analysis
    • The report output from Single Cell V(D)J-Seq Analysis
    • The clonotype identification algorithm
  • Filter Cell Clonotypes
  • Combine Cell Clonotypes
  • Compare Cell Clonotypes
    • The output of Compare Cell Clonotypes
  • Convert Clonotypes to Cell Annotations
  • The Cell Clonotypes element
    • Primary and secondary clonotypes
    • Cell Clonotypes tables
    • Cell Clonotypes alignments
    • Cell Clonotypes Sankey plot
• Noise reduction through feature selection and dimensionality reduction
  • Feature selection and dimensionality reduction
    • Calculation of estimated biological variation
• Cell Annotation
  • Cluster Single Cell Data
    • The Cluster Single Cell Data algorithm
  • Create Heat Map for Cell Abundance
    • The output of Create Heat Map for Cell Abundance
  • Create Cell Annotations from Hashtags
  • The Cell Annotations element
  • The Cell Clusters element
• Dimensionality reduction
  • UMAP for Single Cell
  • tSNE for Single Cell
• Single cell low-dimensional plots functionality
  • Manual annotation
  • Visualizing different types of matrices
  • Create Subset
  • Extract to Table
  • Launching of Create Expression Plot
  • Launching of Create Heat Map for Cell Abundance
  • Launching of Differential Accessibility for Single Cell
  • Launching of Differential Expression for Single Cell
  • Launching of Differential Velocity for Single Cell
  • Launching of Score Velocity Genes
• Utility tools
  • Combine Cell Annotations
  • Update Cell Annotations
  • Convert Metadata to Cell Annotations
  • Combine Cell Clusters
  • Update Cell Clusters
  • Add Information to Plot
  • Update Single Cell Sample Name
  • Split by Cell
• Single cell template workflows from reads
  • Expression Analysis from Reads
    • Configuring the batch units for Expression Analysis from Reads
    • Output from Expression Analysis from Reads
  • Chromatin Accessibility Analysis from Reads
    • Configuring the batch units for Chromatin Accessibility Analysis from Reads
    • Output from Chromatin Accessibility Analysis from Reads
  • Chromatin Accessibility and Expression Analysis from Reads
    • Configuring the batch units for Chromatin Accessibility and Expression Analysis from Reads
    • Output from Chromatin Accessibility and Expression Analysis from Reads
    • Importing reads
  • Immune Repertoire Analysis from Reads (10xV(D)J)
    • Output from Immune Repertoire Analysis from Reads (10xV(D)J)
  • Immune Repertoire and Expression Analysis from Reads (10xV(D)J)
    • Configuring the batch units for Immune Repertoire and Expression Analysis from Reads (10xV(D)J)
    • Output from Immune Repertoire and Expression Analysis from Reads (10xV(D)J)
• Single cell template workflows from imported data
  • Expression Analysis from Matrix
    • Output from Expression Analysis from Matrix
  • Chromatin Accessibility and Expression Analysis from Matrix
    • Output of Chromatin Accessibility and Expression Analysis from Matrix
  • Immune Repertoire and Expression Analysis from Clonotypes and Matrix
    • Output from Immune Repertoire and Expression Analysis from Clonotypes and Matrix
• Bibliography

The output of Normalize Single Cell Data

Normalize Single Cell Data produces the following outputs:

• A normalized Expression Matrix () / (). It is not possible to see the normalized expressions directly in the matrix. Downstream tools automatically use them when available.
• Optionally, a Report () providing a summary and diagnostic plots.

The report includes the following sections.

Summary

Contains a table with:

• Target genes. The number of genes for which the model was fitted.
• Not fitted genes. The number of genes for which the model fit failed.

See The Normalize Single Cell Data algorithm for more details.

Variation of normalized expressions

Contains the following subsections:

• Variance of Pearson residual expression. A plot containing the residual variance. Since most genes are expected to have relatively stable expression across all cells, the residual variance is expected to be ~1 for the majority of genes. The plot is useful for determining if the model is inappropriate for the data. This can happen in several ways:
  • The distribution of expressions is not negative binomial (NB). The NB model is most appropriate for UMI data. Figure 7.26 shows a dataset without UMIs and with very deep sequencing. Even if the NB model might not be a perfect fit, normalization may still be beneficial, but it is important to check whether the plot indicates problems with the data.

    Usually this is not a problem, as the negative binomial (NB) model is quite flexible. but if it doesn't fit an NB it might still be good enough.

    (I don't think it matters, but if it doesn't fit an NB it might still be good enough. You can use the wrong distribution and get satisfactory results e.g. normal distribution for people's heights

    
    Figure 7.26: Residual variance plot for data without UMIs, which is not modeled well by the negative binomial distribution.

  • The normalization is over-parameterized. The model includes a term for each batch, which is estimated from the data. The fewer the cells in a batch, the less accurate the estimation will be.

  • The model is under-parameterized. Figure 7.27 illustrates the effect of batch correcting data containing a mixture of multiple protocols. Without batch correction (plot to the left), the points deviate more noticeably from the expected line. In contrast, for the batch corrected data (plot to the right), the points are more tightly clustered around the expected line.

    
    Figure 7.27: Residual variance plots for data containing a mixture of several protocols. Left: No batch correction is performed and the model is under-parameterized. Right: Batch correction is performed.

• Highly variable genes. A table with the most highly variable genes found in the data after normalization. This list should typically be enriched with marker genes representing the cell types present in the data.

  It can also provide insights into potential issues. For example:

  • If multiple samples were supplied as input, and the list contains many rRNA genes, it could indicate that some samples had higher rRNA content. Further investigation might reveal that the samples were prepared by two different investigators, suggesting the need to include this as a batch factor.
  • If batch correcting for cell cycle effects, and the list contains many genes known to have a role in the cell cycle, it could indicate that batch correction was not successful.

Fitted values

The remaining sections of the report contain the fitted values for each model term: 'Intercept', 'Log10(theta)' (figure 7.28), and a term for each batch effect (figure 7.29).

These sections can be useful for assessing whether the model is likely over-parameterized, as it easy to see the number of terms. Each section contains a plot with the fitted values and, where relevant, the regularized trend.

The plots for the batch effect terms show the fold change for the corresponding batch relative to the baseline (figure 7.29). It can be useful to inspect the fold change of individuals genes. For this, open the plot outside the report by double-clicking on it, switch to the table view (), and sort the table by fold change. For example, when correcting for cell cycle effects, genes involved in the cell cycle are expected to exhibit large positive and negative fold changes.

Figure 7.28: Fitted dispersion . The red line is the regularized trend, providing a more robust estimate of the dispersion. The expression of some genes is consistent with a Poisson distribution, seen here by the band of genes at .

Figure 7.29: Batch effect term. The y-axis shows the natural logarithm of the fold change in the CL sample relative to the baseline SM2 sample. The large fold changes, concentrated in two bands at y=20 and y=-20, correspond to genes that are not expressed in at least one of the samples.

---