Normalization and Subsampling

Overview

This vignette describes how PRIME handles normalization across libraries and subsampling for saturation analyses in CTSS datasets.

The examples use a CTSS RangedSummarizedExperiment object bundled with PRIME. See vignette("ctss-processing") for how to build this object from BigWig files.

Setup

library(PRIME)
library(SummarizedExperiment)

ctss <- readRDS(system.file("extdata", "ctss.rds", package = "PRIME"))

Size factor normalization

PRIME provides two size-factor-based normalization functions:

# Compute size factors (e.g., from total tags or an external method)
size_factors <- colData(ctss)$totalTags

# Normalize raw counts by size factors
ctss_norm <- PRIME::normalizeBySizeFactors(
    ctss,
    sizeFactors  = size_factors,
    inputAssay   = "counts",
    outputAssay  = "normalized"
)

# Normalize to TPM scaled by size factors
ctss_norm <- PRIME::TPMnormalizeBySizeFactors(
    ctss,
    sizeFactors  = size_factors,
    inputAssay   = "counts",
    outputAssay  = "normalizedTPM"
)

Subsampling

Subsampling is used to equalize library depth before comparative analyses and for sequencing saturation experiments.

Subsample to an absolute target depth

target_depth <- min(colData(ctss)$totalTags)
ctss_sub <- PRIME::subsampleTarget(ctss, target = target_depth)

Subsample to a proportion of the library

ctss_prop <- PRIME::subsampleProportion(ctss, proportion = 0.5)

Batch-aware workflows

The bundled example data do not contain batch labels, so batch normalization is not run here. For multi-batch datasets, PRIME provides PRIME::calcBatchSupport() to summarize support across batches and PRIME::conditionalNormalize() to normalize within batches. Use a batch vector with one entry per sample, for example batch = colData(ctss)$Batch, when calling these functions.

Export normalized BigWig files

After normalization, export per-sample BigWig files for visualization or downstream tools:

# Create the directory if it doesn't already exist
if (!dir.exists("bw_output")) {
  dir.create("bw_output", recursive = TRUE)
}

PRIME::writeBw(
    ctss,
    replicates = "all",
    dir        = "bw_output/"
)