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

This vignette explores the design of the summary method for HermesData objects.

Idea

The user would like to get some high level summary statistics for a given HermesData object.

Inspiration sources

Default summary method

By default we get the following summary:

object <- hermes_data
summary(object)

DESeq2

Let's see what the DESeq2 package provides.

library(DESeq2)
countData <- matrix(1:100,ncol=4)
condition <- factor(c("A","A","B","B"))
dds <- DESeqDataSetFromMatrix(countData, DataFrame(condition), ~ condition)
class(dds)
summary(dds)

So nothing in addition.

Vector

Both of these actually just inherit through SummarizedExperiment from S4Vectors::Vector with the corresponding method:

getMethod("summary", "Vector")
S4Vectors:::.Vector_summary

Ideas

Specialized text

As a minimum we could specialize a bit the text of the summary but otherwise keep it similar like the above.

.summary1 <- function (object) 
{
    n_genes <- nrow(object)
    n_samples <- ncol(object)
    n_feature_cols <- ncol(rowData(object))
    n_sample_cols <- ncol(colData(object))

    paste(
      classNameForDisplay(object), "object with", 
      n_samples, "samples of", n_genes, "genes, described by",
      n_feature_cols, "feature data columns and", n_sample_cols, "sample data columns."
    )
}

.summary1(object)

rhd <- HermesData(as(summarized_experiment, "RangedSummarizedExperiment"))
.summary1(rhd)

Add more information

We can add more information. Now for that, it is better that we proceed in steps. 1. have a summary method that creates a HermesDataSummary object 1. have a show method for HermesDataSummary that produces the console output

That way we separate the extraction / calculation of information from printing, and the user could also access summary information directly.

.HermesDataSummary <- setClass(
  Class = "HermesDataSummary",
  slots = c(
    class_name = "character",
    n_genes = "integer",
    n_samples = "integer",
    additional_feature_cols = "character",
    additional_sample_cols = "character",
    no_qc_flags_filled = "logical",
    genes_fail = "character",
    samples_fail = "character",
    lib_sizes = "numeric",
    assay_names = "character"
  )
)

.summary2 <- function(object) 
{
  rd <- rowData(object)
  additional_feature_cols <- setdiff(
    names(rd),
    union(.row_data_non_empty_cols, .row_data_additional_cols)
  )
  genes_fail <- rownames(object)[which(rd$LowExpressionFlag)]
  cd <- colData(object)
  additional_sample_cols <- setdiff(
    names(cd),
    union(.col_data_non_empty_cols, .col_data_additional_cols)
  )
  samples_fail <- colnames(object)[which(cd$TechnicalFailureFlag | cd$LowDepthFlag)]
  no_qc_flags_filled <- is.null(metadata(object)$control_quality) &&
    all_na(rd$LowExpressionFlag) &&
    all_na(cd$TechnicalFailureFlag) &&
    all_na(cd$LowDepthFlag)
  .HermesDataSummary(
    class_name = classNameForDisplay(object),
    n_genes = nrow(object),
    n_samples = ncol(object),
    additional_feature_cols = additional_feature_cols,
    additional_sample_cols = additional_sample_cols,
    no_qc_flags_filled = no_qc_flags_filled,
    genes_fail = genes_fail,
    samples_fail = samples_fail,
    lib_sizes = colSums(counts(object)),
    assay_names = assayNames(object)
  )
}

.summary2(object)

We can define this as the summary method:

setMethod(
  f = "summary",
  signature = c(object = "AnyHermesData"),
  definition = .summary2
)

Therefore we can use it:

summary(object)

Now we can write the show method:

setMethod(
  f = "show",
  signature = c(object = "HermesDataSummary"),
  definition = function(object) {
    cat(
      object@class_name, "object with", 
      object@n_samples, "samples of", object@n_genes, "genes."
    )
    # etc. - more here to come during production. 
  }
)

So we can use it - implicitly or explicitly:

summary(object)
res <- summary(object)
res
str(res)

insightsengineering/hermes documentation built on Dec. 15, 2024, 8:07 a.m.