expressionCor()

Adrià Mitjavila Ventura

Source: vignettes/14-expressionCor.Rmd

Load packages

Run expressionCor()

Required input

The input for expressionCor() is a data frame with a Geneid column and several columns with expression values of each sample. It is required that the values in the Geneid column are unique.

expr <- read.delim("../testdata/tpm.tsv")
expr[1:5,1:5]
##        Geneid cond1_rep1 cond1_rep2 cond2_rep1 cond2_rep2
## 1     DDX11L1  0.0000000   0.000000   0.000000   0.000000
## 2      WASH7P  8.6190283   5.066649   5.182149   1.334332
## 3   MIR6859-1 76.1065829   3.728237   0.000000   2.945561
## 4 MIR1302-2HG  0.2112001   0.000000   0.000000   0.000000
## 5   MIR1302-2  0.0000000   0.000000   0.000000   0.000000

Minimum run 

expressionCor(df = expr)

Groups and facets

expressionCor() allows to use groups, which then will be used to separate the correlation plots into facets. To enable this, set the parameter by_groups to TRUE. This option requires an extra column named group in which each row will have a determined value.

expr2 <- expr %>% dplyr::mutate(group = c(rep(c("group1", "group2", "group3"), nrow(expr)/3), "group1", "group2"))
head(expr2)
##        Geneid cond1_rep1 cond1_rep2 cond2_rep1 cond2_rep2 cond3_rep1 cond3_rep2
## 1     DDX11L1  0.0000000   0.000000  0.0000000   0.000000  0.0000000   0.000000
## 2      WASH7P  8.6190283   5.066649  5.1821493   1.334332  1.9376567   4.092955
## 3   MIR6859-1 76.1065829   3.728237  0.0000000   2.945561  0.0000000   3.388224
## 4 MIR1302-2HG  0.2112001   0.000000  0.0000000   0.000000  0.0000000   0.000000
## 5   MIR1302-2  0.0000000   0.000000  0.0000000   0.000000  0.0000000   0.000000
## 6     FAM138A  0.0000000   0.000000  0.1852678   0.000000  0.2386085   0.000000
##    group
## 1 group1
## 2 group2
## 3 group3
## 4 group1
## 5 group2
## 6 group3
expressionCor(df = expr2, samples = colnames(expr2)[-1], by_groups = T)

Options

In expressionCor(), the default is to compute the Pearson correlation. Nevertheless, the correlation method can be changed to Spearman or Kendall correlations by changing the corr_method argument to "spearman" or "kendall".

expressionCor(df = expr, corr_method = "pearson") + labs(title = "Pearson correlation (default)") +
expressionCor(df = expr, corr_method = "spearman") + labs(title = "Spearman correlation")

Also, the genes and/or samples used to compute the correlation values can be subset using the genes and samples parameters.

expressionCor(df = expr, genes = c("WASH7P", "FAM138A", "AL627309.1")) + labs(title = "Subset of genes") +
expressionCor(df = expr, samples = c("cond3_rep2", "cond1_rep1", "cond2_rep1")) + labs(title = "Subset of samples")

Customization

Heatmap

expressionCor() allows an easy customization of the heatmap:

  • Plot type:

    • by default, expressionCor() plots only the lower part of the matrix (since it is symmetrical), including the diagonal.
    • setting the parameter plot_type to "upper" (plots the upper part of the matrix) or "full" (plots the whole matrix), the plot type can be changed.
expressionCor(df = expr, plot_type = "lower") + labs(title = "Lower plot (default)") +
expressionCor(df = expr, plot_type = "upper") + labs(title = "Upper plot") +
expressionCor(df = expr, plot_type = "full") + labs(title = "Full plot")

  • Diagonal:

    • expressionCor() plots the diagonal by default.
    • by setting plot_diagonal = F, the diagonal is not drawn.
expressionCor(df = expr, plot_diagonal = F)

  • Correlation values:

    • expressionCor() allows changing the coefficients color with the parameter coeffs_color, which is black by default.
    • the correlation coefficients can be removed from the plot if coeffs_color is set to NULL.
    • the parameter coeffs_size allows the customization of the coefficients size, which is 4 by default.
expressionCor(df = expr, coeffs_color = "Black") +
expressionCor(df = expr, coeffs_color = "Red", coeffs_size = 6)

  • Colors:

    • the colors of the plot can be changed by setting the parameter plot_colors with a character vector with 3 valid R colors.
    • the values from plot_colors will be passed to scale_fill_gradient2() to set the color of the limits and the midpoint.
expressionCor(expr, plot_colors = c("blue", "yellow", "red"))

  • Size:

    • the size of the heatmap (both with and height) can be changed with the parameter plot_size which, by default is 85 (millimeters).
    • changing the plot size, will automatically change the legend size to adjust it to the height (if the legend is vertical) or width (if the legend is horizontal).
expressionCor(df = expr, plot_size = 50)

  • Title, subtitle, caption:

    • the parameters plot_title, plot_subtitle and plot_caption allow to specify the title, subtitle and caption of the plot, respectively.
    • also, the text of the title, subtitle, caption can be formatted with the parameters, title_hjust, title_face, title_size, subtitle_face, subtitle_size, caption_size.
expressionCor(df = expr, plot_title = "Title", plot_subtitle = "Subtitle", plot_caption = "Caption")

  • Axes:

    • the text in the axes can be formatted with axis_text_color and axis_text_size 
expressionCor(df = expr,axis_text_color = "blue", axis_text_size = 7)

  • Cell border:

    • the parameter cell_border allows to set the color of the cell border which, by default is gray. It can be set to a valid R color or to a character “NA”, which removes the border.
expressionCor(df = expr) +
expressionCor(df = expr, cell_border = "Red") +
expressionCor(df = expr, cell_border = "NA")

Legend

expressionCor() allows an easy customization of the legend:

  • Legend limits:

    • by setting the parameters legend_limits and legend_breaks_by, you can change the legend scale which, by default, goes from -1 to 1.
    • the legend midpoint is automatically set to the mean of minimum and maximum values.
expressionCor(df = expr, legend_limits = c(0.9,1), legend_breaks_by = 0.05)

  • Legend title:

    • change the legend_title parameter which, by default, is set to the correlation method used (e.g. if the correlation is Pearson’s, the title is Pearson correlation).
expressionCor(df = expr, legend_title  = "Title of the legend")

  • Position of the legend:

    • the legend_pos parameter allows to position the legend to the left, right, top or bottom.
(expressionCor(df = expr, legend_pos = "left", legend_title = "Legend at the left") +
expressionCor(df = expr, legend_pos = "right", legend_title = "Legend at the right")) /
(expressionCor(df = expr, legend_pos = "top", legend_title = "Legend at the top") +
expressionCor(df = expr, legend_pos = "bottom", legend_title = "Legend at the bottom"))

  • Legend size: by setting the parameter legend_size with a numerical vector of length 2, the size of the legend can be changed.

    • If the legend is vertical (i.e., at the left or at the right), the first number is the width and the second is the height of the legend, in millimeter.
    • If the legend is horizontal (i.e., at the top or at the bottom), the first element is the height and the second is the width, in millimiters.
    • By default, the legend size equals to the heatmap size.
expressionCor(df = expr, legend_size = c(5,50)) +
expressionCor(df = expr, legend_pos = "bottom", legend_size = c(5,50))