Complete preprocessing pipeline for tAge analysis

This function performs a complete preprocessing pipeline for transcriptomic age analysis, including gene filtering, normalization, transformation, scaling, control subtraction, and gene ID conversion. It returns multiple versions of the processed data suitable for different analysis approaches.

Usage

tAge_preprocessing(
  eset,
  species = "mouse",
  gene_mapping_type = "auto",
  verbose = TRUE,
  control_group_column = NULL,
  control_group_label = NULL,
  count_threshold = 10,
  percent_threshold = 20,
  split_by = NULL
)

Arguments

eset: An ExpressionSet object containing raw integer counts (no NA, no negative or non-integer values; normalised data are refused).
species: Species of the samples: "mouse", "rat", "human" or "monkey" (see tage_species). Default is "mouse".
gene_mapping_type: Identifier type of the row names: "Ensembl", "Gene.Symbol", "Entrez" or "auto" (default, detected from the gene table; see map_genes).
verbose: Logical indicating whether to print progress messages. Default is TRUE.
control_group_column: Character string specifying the column name in phenoData that contains control group labels. Default is NULL.
control_group_label: Character string specifying the label for control samples. Default is NULL. With split_by, the controls of each stratum are its reference; a stratum without controls is centred on all of its samples, with a warning. A column that does not exist, or a label that no sample carries, is an error.
count_threshold: Numeric threshold for minimum expression count in gene filtering. Default is 10.
percent_threshold: Numeric threshold for minimum percentage of samples that must have expression above count_threshold. Default is 20.
split_by: Character or NULL. Column of the phenoData whose levels (tissues, datasets, cell types) are preprocessed separately. Default NULL.

Value

A list containing six processed ExpressionSet objects:

RLE_normalized: RLE-normalized data
log_transformed: Log-transformed data
scaled: Scaled data with gene ID conversion
scaled_diff: Scaled data with control subtraction and gene ID conversion
yugene: YuGene-normalized data with gene ID conversion
yugene_diff: YuGene-normalized data with control subtraction and gene ID conversion

Details

The clocks were trained on expression centred within each dataset and tissue against matched controls, and that is how they should be applied: with several tissues (or datasets, cell types) in one ExpressionSet, pass split_by so that gene filtering, normalisation and reference centring all happen within each stratum, and the results are combined afterwards. Without split_by the whole object is one stratum.

The species is recorded in every returned ExpressionSet, so predict_tAge does not need it again.

Examples

# Load example data
expr_data <- load_example_expression_data()
meta_data <- load_example_metadata()
eset <- make_ExpressionSet(expr_data, meta_data)
#> ✓ ExpressionSet created successfully
#>   - Number of genes: 57010 
#>   - Number of samples: 24 


# Run complete preprocessing pipeline
processed_data <- tAge_preprocessing(eset, species = "mouse")
#> ✓ Gene filtering completed
#>   - Number of genes before filtering: 57010 
#>   - Number of genes after filtering: 19550 
#>   - Percentage of genes retained: 34.3 %
#> ✓ Gene identifiers: Ensembl (15996 of 19550 row names found in the mouse gene table)
#> calcNormFactors has been renamed to normLibSizes
#> ✓ RLE normalization completed

#> ✓ Log transformation completed

#> ✓ Scaling completed

#> ✓ YuGene normalization completed

#> ✓ Centring on all samples (overall per-gene median; no reference group specified).
#> Warning: NaNs produced

#> ✓ Centring on all samples (overall per-gene median; no reference group specified).


# Two tissues: filter, normalise and centre each on its own wild-type samples
processed_data <- tAge_preprocessing(
  eset, species = "mouse", split_by = "Tissue",
  control_group_column = "Genotype", control_group_label = "WT"
)
#> 
#> === Tissue = Kidney (12 samples) ===
#> ✓ Gene filtering completed
#>   - Number of genes before filtering: 57010 
#>   - Number of genes after filtering: 19374 
#>   - Percentage of genes retained: 34 %
#> ✓ Gene identifiers: Ensembl (15818 of 19374 row names found in the mouse gene table)
#> calcNormFactors has been renamed to normLibSizes
#> ✓ RLE normalization completed

#> ✓ Log transformation completed

#> ✓ Scaling completed

#> ✓ YuGene normalization completed

#> ✓ Control samples found for label 'WT'. Using control group median for subtraction.
#> Warning: NaNs produced

#> ✓ Control samples found for label 'WT'. Using control group median for subtraction.

#> 
#> === Tissue = Skeletal muscle (12 samples) ===
#> ✓ Gene filtering completed
#>   - Number of genes before filtering: 57010 
#>   - Number of genes after filtering: 16423 
#>   - Percentage of genes retained: 28.8 %
#> ✓ Gene identifiers: Ensembl (14100 of 16423 row names found in the mouse gene table)
#> calcNormFactors has been renamed to normLibSizes
#> ✓ RLE normalization completed

#> ✓ Log transformation completed

#> ✓ Scaling completed

#> ✓ YuGene normalization completed

#> ✓ Control samples found for label 'WT'. Using control group median for subtraction.
#> Warning: NaNs produced

#> ✓ Control samples found for label 'WT'. Using control group median for subtraction.