volcano_plot: Volcano plot

In jmw86069/jamma: MA-plots for omics data

Volcano plot

Description

Draw a volcano plot using reasonable default arguments.

Usage

volcano_plot(
  x,
  n = NULL,
  lfc_colname = c("logfc", "log2fold", "log2fc", "lfc", "l2fc", "logratio", "log2ratio"),
  fold_colname = c("fold", "fc", "ratio"),
  fold_cutoff = 1.5,
  fold_max_range = 16,
  fold_min_range = 4,
  sig_colname = c("adj.P.Val", "padj", "adj.pval", "adjp", "P.Value"),
  sig_cutoff = 0.05,
  sig_max_range = 1e-10,
  sig_min_range = 1e-04,
  expr_colname = c("mgm", "groupmean", "mean", "AveExpr", "fkpm", "rpkm", "tpm", "cpm"),
  expr_cutoff = NULL,
  label_colname = c("gene", "symbol", "protein", "probe", "assay"),
  main = "Volcano Plot",
  submain = NULL,
  blockarrow = TRUE,
  blockarrow_colors = c(hit = "#E67739FF", up = "#990000FF", down = "#000099FF"),
  blockarrow_font = 1,
  blockarrow_cex = c(1.2, 1.2),
  blockarrow_label_cex = 1,
  blockarrow_shadowtext = TRUE,
  symmetric_axes = TRUE,
  do_cutoff_caption = TRUE,
  caption_cex = 0.8,
  include_axis_prefix = FALSE,
  n_x_labels = 12,
  n_y_labels = 7,
  xlim = NULL,
  ylim = NULL,
  pt_cex = 0.9,
  pt_pch = 21,
  hit_type = "hits",
  color_set = c(base = "#77777777", up = "#99000088", down = "#00009988", hi =
    "#FFDD55FF", hi_up = "#FFDD55FF", hi_down = "#FFDD55FF"),
  border_set = NULL,
  point_colors = NULL,
  border_colors = NULL,
  abline_color = "#000000AA",
  smooth = TRUE,
  smooth_func = jamba::plotSmoothScatter,
  smooth_ramp = colorRampPalette(c("white", "lightblue", "lightskyblue3", "royalblue",
    "darkblue", "orange", "darkorange1", "orangered2")),
  tophist = FALSE,
  tophist_cutoffs = c("pvalue", "foldchange"),
  tophist_breaks = 100,
  tophist_color = "#000099FF",
  tophist_fraction = 1/3,
  tophist_by = 0.2,
  hi_points = NULL,
  hi_colors = NULL,
  hi_hits = FALSE,
  hi_cex = 1,
  do_both = FALSE,
  label_hits = FALSE,
  add_plot = FALSE,
  xlab = NULL,
  ylab = NULL,
  cex.axis = 1.2,
  mar_min = c(6, 5, 6, 5),
  transFactor = 0.24,
  transformation = function(x) {
     x^transFactor
 },
  nbin = 256,
  verbose = TRUE,
  ...
)

Arguments

x	data.frame that contains statistical results with at least a P-value, and fold change or log2 fold change. It is useful to contain a column with mean expression, and a column with a relevant label.
n	integer indicating the number of subset points to plot for testing purposes.
lfc_colname	character string or vector used to match colnames(x) whose values should be log2 fold changes. A direct match to colnames(x) is performed first, then if no column is found, the values are used as regular expression patterns in order until the first matching colname is found. Note that lfc_colname is used in preference to fold_colname. The colname used will appear as the x-axis label.
fold_colname	character string or vector used to match colnames(x) whose values should be fold changes. Note that if lfc_colname successfully finds a value, the fold_colname is not used. The colname if used will appear as the x-axis label.
fold_cutoff	numeric threshold for values in lfc_colname or fold_cutoff, where normal fold change values at or above fold_cutoff can be considered statistically significant. Note that when lfc_colname is being used, its values are converted to normal fold change before applying this filter.
fold_max_range	numeric indicating the maximum range to display on the x-axis fold change. This argument prevents extremely large fold changes from compressing the useful visible range of the figure.
fold_min_range	numeric indicating the minimum range to display on the x-axis fold change. This argument is useful when fold changes are low and the x-axis range would otherwise be too small to be very useful.
sig_colname	character string or vector used to match colnames(x) whose values should contain P-values of significance. The P-values can be unadjusted (raw) P-values, or adjusted P-values. The P-values are expected not to be -log10() transformed. The colname used will appear as the y-axis label.
sig_cutoff	numeric threshold for values in sig_colname, where values at or below sig_cutoff can be considered statistically significant.
sig_max_range	numeric indicating the maximum range to display on the y-axis significance. This argument prevents extremely small P-values from compressing the useful visible range of the figure.
sig_min_range	numeric indicating the minimum range to display on the y-axis significance. This argument is useful when P-values are not very significant, and you want to make sure the y-axis range shows a minimum amount of the significant region to be visually interpretable in that context.
expr_colname	character string or vector used to match colnames(x) whose values should contain expression mean values. This column is only used when expr_cutoff is defined and is applied to the filter criteria for statistical hits.
expr_cutoff	numeric threshold for values in expr_colname when expr_colname is defined, where values in expr_colname at or above expr_cutoff can be considered statistically significant. This threshold is useful to filter out potential statistical hits whose signal is below a noise signal threshold.
label_colname	character string or vector used to match colnames(x) whose values should contain a useful label, for example gene symbol or assay identifier.
main	character string used as the main title of the figure.
submain	character string used as a sub-title of the figure.
blockarrow	logical indicating whether block arrows should be displayed and used to indicate the number of statistical hits.
blockarrow_colors, blockarrow_font, blockarrow_label, blockarrow_shadowtext	arguments used when blockarrow=TRUE.
symmetric_axes	logical indicating whether the x-axis log fold change range should be symmetric above and below zero.
do_cutoff_caption	logical indicating whether to display text caption with the statistical cutoff values used, and the total number of points displayed.
caption_cex	numeric caption font size adjustment.
include_axis_prefix	logical indicating whether to include a prefix for the x-axis and y-axis labels: x-axis "Change"; y-axis "Significance".
n_x_labels, n_y_labels	integer used by pretty() to determine the approximate number of x-axis and y-axis labels to display, respectively.
xlim, ylim	numeric used to define specific xlim and ylim axis ranges. When NULL the ranges are defined automatically, using fold_min_range,fold_max_range for the x-axis, and sig_min_range,sig_max_range for the y-axis.
pt_cex, pt_pch	numeric used to define point size and shape, used only when individual points are displayed.
hit_type	character string used to label points that meet the statistical cutoffs as "hits", but where it may be useful to indicate the type of entry being tested. For example: hit_type="genes" indicates that each row represents a gene; hit_type="probes" indicates each row represents a probe; hit_type="transcripts" indicates each row represents a transcript.
color_set	character vector of R colors, used only when individual points are display. The names override default values, and may include: "base" - the base color of all points on the plot "up" - the color for up-regulated points that meet all statistical cutoffs to be a "hit". "down" - the color for down-regulated points that meet all cutoffs "hi" - base color for highlighted points, used when hi_points is defined. "hi_up" - color for highlighted up-regulated points. "hi_down" - color for highlighted down-regulated points.
border_set	NULL or character vector of R colors, used to define point border colors such as pch=21 which is a filled circle with border. When border_set=NULL then it is defined by jamba::makeColorDarker(color_set).
point_colors, border_colors	optional character vector of R colors recycled to length nrow(x), used to specify the exact color of each point in x. This argument is useful to colorize certain specific points that may otherwise not meet statistical criteria.
abline_color	character string with R color used to color the abline that indicates the x-axis fold_cutoff value, and y-axis sig_cutoff value.
smooth	logical indicating whether points should be drawn as a smooth scatter plot, using jamba::plotSmoothScatter(). When smooth=FALSE individual points are drawn, using point_colors, or when point_colors is not defined the default is to use color_set to colorize points based upon statistical cutoffs.
smooth_func	function used to plot points when smooth=TRUE, by default jamba::plotSmoothScatter() which has some benefits over default graphics::smoothScatter().
smooth_ramp	character vector of R colors which defines the color gradient to use when smooth=TRUE.
tophist	logical indicating whether to display a histogram at the top of the volcano plot figure.
tophist_cutoffs, tophist_breaks, tophist_fraction, tophist_by	arguments used when tophist=TRUE.
hi_points	character vector indicating points to highlight in the volcano plot, where values should match rownames(x). This argument is useful to highlight a specific subset of points of interest on the figure. Note that hi_points are always rendered as individual points even when smooth=TRUE.
hi_hits	logical indicating whether rows that meet all statistical cutoffs and are considered "hits" should also be treated as hi_points for the purpose of rendering individual points.
hi_cex	numeric size adjustment for highlight points, relative to the size of other points in the figure.
do_both	logical indicating whether to draw both a smooth scatter and individual points on the same figure.
label_hits	logical indicating whether to add a text label for points that are statistical hits.
add_plot	logical indicating whether the plot should be added to an existing plot, or when add_plot=FALSE a new plot is created. This argument is useful to re-run the same volcano plot with alternate parameters, for example to display different subsets of highlighted points.
xlab, ylab	character strings used to specify the exact x-axis label and y-axis label. When either value is NULL the default is to use the relevant colname: x-axis uses either lfc_colname or fold_colname; y-axis uses sig_colname.
cex.axis	numeric adjustment for axis label font sizes.
transformation	function passed to smooth_func used to adjust the visual contrast of the resulting density plot.
nbin	numeric value passed to smooth_func and used by jamba::plotSmoothScatter() to adjust the number of bins used to display the density of points, where a higher value shows more detail, and a lower value shows less detail.
verbose	logical indicating whether to print verbose output. Note that verbose=2 will enable much more verbose output.
...	additional arguments are ignored.
`numeric`	vector used to ensure that each margin size is at least a minimum value, applied to par("mar") via the function pmax().

Details

Draw a volcano plot using a reasonably robust set of default arguments, and with a large number of customization options. The default plot uses smooth scatter plot for much improved display of point density.

This function produces a volcano plot, which consists of change on the x-axis, and significance on the y-axis.

In addition to displaying the volcano plot, this function also displays statistical thresholds, and marks entries as "hits" by up to three conceptual filters:

• "change" - fold change fold_cutoff

• "significant" - statistical P-value sig_cutoff

• "detected" - signal expr_cutoff

If any cutoff is not defined, that filter is ignored.

Change is usually represented using log2 fold changes, and in this case is labeled using normal scale fold change values. The threshold is defined with fold_cutoff using normal space values. The log2 fold change values which have greater magnitude than fold_cutoff are marked "changing".

Significance usually represents adjusted P-value, or raw P-value if necessary. The threshold is defined with sig_cutoff using a P-value below which entries are marked "significant".

Finally, since some statistical criteria also include a minimum level of signal, a threshold expr_cutoff requires an entry to have signal at or above this value to be considered "detected".

The default behavior of volcano_plot() is to render a smooth scatter plot. A smooth scatter plot is much more effective at representing the true point density along the figure, which is one of the primary reasons to produce the plot.

Highlighting points

The argument hi_points can be used to highlight a specific subset of points on the figure, even when smooth=TRUE.

Alternatively, hi_hits=TRUE will render all statistical hits as points, which will appear on top of the smooth scatter plot when smooth=TRUE.

See Also

Other jam plot functions: ggjammaplot()

Examples

n <- 15000;
set.seed(12);
x_lfc <- (rnorm(n) * 1);
x_lfc <- x_lfc^2 * sign(x_lfc);
x_lfc <- x_lfc[order(-abs(x_lfc) + rnorm(n) / 2)];
x_pv <- sort(10^-(rnorm(n)*1.5)^2);
x <- data.frame(
   Gene=paste("gene", seq_len(n)),
   `log2fold Group-Control`=x_lfc,
   `P.Value Group-Control`=x_pv[order(-abs(x_lfc))],
   `mgm Group-Contol`=((rnorm(1500)+5)^2)/5,
   check.names=FALSE);

volcano_plot(x);
volcano_plot(x, expr_cutoff=3);
# volcano_plot(x, mar_min=c(7, 6, 6, 5), blockarrow_cex=1);

# par("mfrow"=c(2, 1));
# volcano_plot(x);
# volcano_plot(x);
# par("mfrow"=c(1, 1));

x[["fold Group-Control"]] <- log2fold_to_fold(x[["log2fold Group-Control"]]);
x[["adj.P.Val Group-Control"]] <- x[["P.Value Group-Control"]];

volcano_plot(x, hi_hits=TRUE);

jmw86069/jamma documentation built on Oct. 11, 2024, 7:08 a.m.