diff_expression: Differential Expression Analysis

In insightsengineering/hermes: Preprocessing, analyzing, and reporting of RNA-seq data

Differential Expression Analysis

Description

The diff_expression() function performs differential expression analysis using a method of preference.

A corresponding autoplot() method is visualizing the results as a volcano plot.

Usage

diff_expression(object, group, method = c("voom", "deseq2"), ...)

## S4 method for signature 'HermesDataDiffExpr'
autoplot(object, adj_p_val_thresh = 0.05, log2_fc_thresh = 2.5)

Arguments

object	(AnyHermesData) input. Note that this function only uses the original counts for analysis, so this does not need to be normalized.
group	(string) name of factor variable with 2 levels in colData(object). These 2 levels will be compared in the differential expression analysis.
method	(string) method for differential expression analysis, see details below.
...	additional arguments passed to the helper function associated with the selected method.
adj_p_val_thresh	(proportion) threshold on the adjusted p-values (y-axis) to flag significance.
log2_fc_thresh	(number) threshold on the absolute log2 fold-change (x-axis) to flag up- or down-regulation of transcription.

Details

Possible method choices are:

• voom: uses limma::voom(), see h_diff_expr_voom() for details.

• deseq2: uses DESeq2::DESeq(), see h_diff_expr_deseq2() for details.

Value

A HermesDataDiffExpr object which is a data frame with the following columns for each gene in the HermesData object:

• log2_fc (the estimate of the log2 fold change between the 2 levels of the provided factor)

• stat (the test statistic, which one depends on the method used)

• p_val (the raw p-value)

• adj_p_val (the multiplicity adjusted p-value value)

Functions

• autoplot(HermesDataDiffExpr): generates a volcano plot for a HermesDataDiffExpr object.

Note

• We provide the df_cols_to_factor() utility function that makes it easy to convert the colData() character and logical variables to factors, so that they can be subsequently used as group inputs. See the example.

• In order to avoid a warning when using deseq2, it can be necessary to specify fitType = "local" as additional argument. This could e.g. be the case when only few samples are present in which case the default parametric dispersions estimation will not work.

Examples

object <- hermes_data %>%
  add_quality_flags() %>%
  filter()

# Convert character and logical to factor variables in `colData`,
# including the below used `group` variable.
colData(object) <- df_cols_to_factor(colData(object))
res1 <- diff_expression(object, group = "SEX", method = "voom")
head(res1)
res2 <- diff_expression(object, group = "SEX", method = "deseq2")
head(res2)

# Pass method arguments to the internally used helper functions.
res3 <- diff_expression(object, group = "SEX", method = "voom", robust = TRUE, trend = TRUE)
head(res3)
res4 <- diff_expression(object, group = "SEX", method = "deseq2", fitType = "local")
head(res4)

# Create the corresponding volcano plots.
autoplot(res1)
autoplot(res3)

insightsengineering/hermes documentation built on March 11, 2024, 11:04 p.m.