imputeIntensity: Imputation methods

In HimesGroup/qmtools: Quantitative Metabolomics Data Processing Tools

Imputation methods

Description

Performs a variety of data imputation methods on a matrix-like object or SummarizedExperiment object. The methods include k-Nearest Neighbors (kNN), Random Forest (RF), and many others from the MsCoreUtils::impute_matrix. See the details below.

Usage

## S4 method for signature 'ANY'
imputeIntensity(
  x,
  method = c("knn", "rf", "bpca", "QRILC", "MLE", "MinDet", "MinProb", "min", "zero",
    "mixed", "nbavg", "with", "none"),
  ...
)

## S4 method for signature 'SummarizedExperiment'
imputeIntensity(
  x,
  method = c("knn", "rf", "bpca", "QRILC", "MLE", "MinDet", "MinProb", "min", "zero",
    "mixed", "nbavg", "with", "none"),
  i,
  name,
  ...
)

Arguments

x	A matrix-like object or SummarizedExperiment object.
method	A string specifying which imputation method to use.
...	Arguments passed to a specific imputation method.
i	A string or integer value specifying which assay values to use when x is a SummarizedExperiment object.
name	A string specifying the name to be used to store the imputed intensities in x when x is a SummarizedExperiment object. If not specified, a matrix containing the imputed intensities is returned.

Details

The method argument can be one of "knn", "rf", "bpca", "QRILC", "MLE", "MinDet", "MinProb", "min", "zero", "mixed", "nbavg", "with", "none". Please choose one that best describes the nature of missing data. While this function provides several simple imputation methods, they may only work under restrictive assumptions.

• "knn" performs kNN imputation based on the Gower distance or Euclidean distance. See imputeKNN for details.

• "rf" performs random forest imputation using the missForest::missForest, as described in Stekhoven D. J., & Buehlmann, P. (2012). This method is not sensitive to monotonic transformations of the intensity matrix.

• For the other method arguments, please refer to the MsCoreUtils::impute_matrix. Briefly,

  • "bpca": Bayesian PCA missing value imputation.

  • "QRILC": Quantile regression approach for the imputation of left-censored missing data.

  • "MLE": Maximum likelihood-based imputation.

  • "MinDet": Deterministic minimal value approach for the imputation of left-censored data.

  • "MinProb": Stochastic minimal value approach for the imputation of left-censored data.

  • "min": Replace the missing values with the smallest non-missing value in the data.

  • "zero": Replace the missing values with 0.

  • "mixed": Mixed imputation applying two methods.

  • "nbavg": Average neighbour imputation for fractions collected along a fractionation/separation gradient.

  • "with": Replace the missing values with a user-provided value.

  • "none": Reserved for the "mixed" method.

Value

A matrix or SummarizedExperiment object of the same dimension as x containing the imputed intensities.

References

Laurent Gatto, Johannes Rainer and Sebastian Gibb (2021). MsCoreUtils: Core Utils for Mass Spectrometry Data. R package version 1.4.0. https://github.com/RforMassSpectrometry/MsCoreUtils

Stekhoven D. J., & Buehlmann, P. (2012). MissForest - non-parametric missing value imputation for mixed-type data. Bioinformatics, 28(1), 112-118.

See Also

See imputeKNN, missForest::missForest, and MsCoreUtils::impute_matrix for the underlying functions that do work.

Examples

data(faahko_se)

## SummarizedExperiment object
se <- imputeIntensity(faahko_se, i = "raw", name = "imp1", method = "knn")
assayNames(se)

## Matrix
m <- assay(faahko_se, i = "raw")
imputeIntensity(m, method = "min")

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