removeICC: Feature Filtering based on ICC

In HimesGroup/qmtools: Quantitative Metabolomics Data Processing Tools

Feature Filtering based on ICC

Description

Removes Features based on a intraclass correlation coefficient (ICC) using the data matrix where rows represent features and columns represent samples. For each feature, ICC will be calculated using both biological and QC samples to identify how much of the total variation is explained by biological variability, as described in Schiffman, C et al (2019). Informative features are expected to have relatively high variability across the biological samples, compared to QC replicates. Features with an ICC below a cut-off will be removed.

Usage

removeICC(
  x,
  qc_samples,
  bio_samples = setdiff(colnames(x), qc_samples),
  cut = 0.4
)

Arguments

x	A matrix-like object.
qc_samples	A vector of sample names or column indices specifying QC samples for the calculation of ICC. Must be a subset of colnames(x) if it is a character vector.
bio_samples	A vector of sample names or column indices specifying biological samples for the calculation of ICC. Must be a subset of colnames(x) if it is a character vector.
cut	A numeric value between 0 and 1 specifying a ICC cut-off to retain a feature.

Value

A matrix containing the filtered features.

References

Schiffman, C., Petrick, L., Perttula, K. et al. Filtering procedures for untargeted LC-MS metabolomics data. BMC Bioinformatics 20, 334 (2019). https://doi.org/10.1186/s12859-019-2871-9

See Also

See removeFeatures that provides a SummarizedExperiment-friendly wrapper for this function.

Examples

set.seed(1e7)

m_bio_1 <- matrix(rlnorm(600, sdlog = 1), ncol = 20)
m_bio_2 <- matrix(rlnorm(200, sdlog = 0.3), ncol = 20)
m_bio <- rbind(m_bio_1, m_bio_2)
m_qc <- matrix(rlnorm(400, sdlog = 0.25), ncol = 10)
m <- cbind(m_bio, m_qc)
colnames(m) <- c(paste0("S", seq_len(20)), paste0("Q", seq_len(10)))

removeICC(m, qc_samples = paste0("Q", seq_len(10)),
          bio_samples = paste0("S", seq_len(20)))

HimesGroup/qmtools documentation built on April 16, 2023, 8 p.m.