Skip to contents

Process Gene Expression Data

Source: R/process_data.R
process_data.Rd

Performs pre-processing, missing value imputation, filtering, and transformation on gene expression count data.

Usage

process_data(
  data,
  missing_value = "geneMedian",
  min_cpm = 0.5,
  n_min_samples_count = 1,
  counts_transform = 0,
  counts_log_start = 4
)

Arguments

data

A numeric matrix or data frame of gene expression counts.

missing_value

Character. Method to handle missing values. Options:

  • "geneMedian": Impute using the median expression of the gene across samples.

  • "treatAsZero": Replace NAs with 0.

  • "groupMedian": Impute using the median of the sample group (detected from colnames).

min_cpm

Numeric. Minimum counts per million threshold for filtering genes.

n_min_samples_count

Numeric. Minimum number of samples that must meet the min_cpm threshold for a gene to be retained.

counts_transform

Integer. Method for data transformation:

  • 0 = None (Raw Counts).

  • 1 = log2(CPM + counts_log_start)

  • 2 = Variance Stabilizing Transformation (VST) via DESeq2.

  • 3 = Regularized Log (rlog) via DESeq2.

counts_log_start

Numeric. Constant added to counts before log transformation. Usually between 1 and 10. Higher values reduce noise but decrease sensitivity (used when counts_transform = 1).

Value

The processed and transformed data matrix.

Examples

# Check example data
summary(pathdb::hypoxia_reads)
#>  MRS1220_hypoxia_rep1 MRS1220_hypoxia_rep2 vehicle_hypoxia_rep1
#>  Min.   :     0.0     Min.   :     0.0     Min.   :     0      
#>  1st Qu.:     0.0     1st Qu.:     0.0     1st Qu.:     0      
#>  Median :     0.0     Median :     1.0     Median :     1      
#>  Mean   :   554.5     Mean   :   554.9     Mean   :   540      
#>  3rd Qu.:   128.0     3rd Qu.:   131.0     3rd Qu.:   124      
#>  Max.   :295557.0     Max.   :301342.0     Max.   :328799      
#>  vehicle_hypoxia_rep2
#>  Min.   :     0      
#>  1st Qu.:     0      
#>  Median :     0      
#>  Mean   :   637      
#>  3rd Qu.:   128      
#>  Max.   :379017      
nrow(pathdb::hypoxia_reads)
#> [1] 35238

# YOU decide how your data is transformed.
# Here, we want to:
# Replace missing values with median
# Set minimum counts-per-million of 0.4
# Meet CPM threshold in 2 samples
# Keep raw counts

hypox_filtered <- process_data(data = pathdb::hypoxia_reads,
                               missing_value = "geneMedian",
                               min_cpm = 0.4,
                               n_min_samples_count = 2,
                               counts_transform = 0)

# Check filtered data
summary(hypox_filtered)
#>  MRS1220_hypoxia_rep1 MRS1220_hypoxia_rep2 vehicle_hypoxia_rep1
#>  Min.   :     0       Min.   :     0.00    Min.   :     0      
#>  1st Qu.:    70       1st Qu.:    72.25    1st Qu.:    67      
#>  Median :   284       Median :   283.00    Median :   269      
#>  Mean   :  1451       Mean   :  1452.09    Mean   :  1413      
#>  3rd Qu.:   937       3rd Qu.:   920.00    3rd Qu.:   886      
#>  Max.   :295557       Max.   :301342.00    Max.   :328799      
#>  vehicle_hypoxia_rep2
#>  Min.   :     0      
#>  1st Qu.:    70      
#>  Median :   289      
#>  Mean   :  1667      
#>  3rd Qu.:  1012      
#>  Max.   :379017      
nrow(hypox_filtered)
#> [1] 13458