Plot Regression Results on Principal Components

The S3 plot method generates plots to visualize the results of regression analyses performed on principal components (PCs) against cell types or dataset origin (query vs. reference).

This function performs linear regression of a covariate of interest onto one or more principal components, based on the data in a SingleCellExperiment object.

# S3 method for class 'regressPCObject'
plot(x, plot_type = c("r_squared", "p-value"), alpha = 0.05, ...)

regressPC(
  reference_data,
  query_data = NULL,
  ref_cell_type_col,
  query_cell_type_col = NULL,
  cell_types = NULL,
  pc_subset = 1:10,
  adjust_method = c("BH", "holm", "hochberg", "hommel", "bonferroni", "BY", "fdr",
    "none"),
  assay_name = "logcounts"
)

Arguments

x

An object of class regressPC containing the output of the regressPC function

plot_type

Type of plot to generate. Options are "r_squared" and "p-value". Default is "r-squared".

alpha

Significance threshold p-values of coefficients. Default is 0.05.

...

Additional arguments to be passed to the plotting functions.

reference_data

A SingleCellExperiment object containing numeric expression matrix for the reference cells.

query_data

A SingleCellExperiment object containing numeric expression matrix for the query cells. If NULL, the PC scores are regressed against the cell types of the reference data.

ref_cell_type_col

The column name in the colData of reference_data that identifies the cell types.

query_cell_type_col

The column name in the colData of query_data that identifies the cell types.

cell_types

A character vector specifying the cell types to include in the plot. If NULL, all cell types are included.

pc_subset

A numeric vector specifying which principal components to include in the plot. Default is PC1 to PC5.

adjust_method

A character string specifying the method to adjust the p-values. Options include "BH", "holm", "hochberg", "hommel", "bonferroni", "BY", "fdr", or "none". Default is "BH" (Benjamini-Hochberg). Default is "BH".

assay_name

Name of the assay on which to perform computations. Default is "logcounts".

Value

The S3 plot method returns a ggplot object representing the specified plot type.

A list containing

• summaries of the linear regression models for each specified principal component,

• the corresponding R-squared (R2) values,

• the variance contributions for each principal component, and

• the total variance explained.

Details

The S3 plot method generates, depending on the specified plot type, either the R-squared values or p-values resulting from the regression of principal components onto cell types or dataset origin (query vs. reference). For cell type regression, the plots show how well each PC correlates with different cell types. For dataset regression, the plots compare the PCs between query and reference datasets.

Principal component regression, derived from PCA, can be used to quantify the variance explained by a covariate of interest. Applications for single-cell analysis include quantification of batch effects, assessing clustering homogeneity, and evaluating alignment of query and reference datasets in cell type annotation settings.

Briefly, the \(R^2\) is calculated from a linear regression of the covariate \(B\) of interest onto each principal component. The variance contribution of the covariate effect per principal component is then calculated as the product of the variance explained by the i-th principal component (PC) and the corresponding \(R^2(PC_i | B)\). The sum across all variance contributions by the covariate effects in all principal components gives the total variance explained by the covariate as follows:

$$Var(C|B) = \sum_{i=1}^G \text{Var}(C|PC_i) \times R^2(PC_i | B)$$

where, \(\text{Var}(C \mid PC_i)\) is the variance of the data matrix \(C\) explained by the i-th principal component. See references for details.

References

Luecken et al. Benchmarking atlas-level data integration in single-cell genomics. Nature Methods, 19:41-50, 2022.

See also

regressPC

plot.regressPCObject

Author

Anthony Christidis, anthony-alexander_christidis@hms.harvard.edu

Examples

# Load data
data("reference_data")
data("query_data")

# Plot the PC data (no query data)
regress_res <- regressPC(reference_data = reference_data,
                         ref_cell_type_col = "expert_annotation",
                         cell_types = c("CD4", "CD8", "B_and_plasma", "Myeloid"),
                         pc_subset = 1:15)
# Plot results
plot(regress_res, plot_type = "r_squared")

plot(regress_res, plot_type = "p-value")


# Plot the PC data (with query data)
regress_res <- regressPC(reference_data = reference_data,
                         query_data = query_data,
                         ref_cell_type_col = "expert_annotation",
                         query_cell_type_col = "SingleR_annotation",
                         cell_types = c("CD4", "CD8", "B_and_plasma", "Myeloid"),
                         pc_subset = 1:15)
# Plot results
plot(regress_res, plot_type = "r_squared")

plot(regress_res, plot_type = "p-value")