reducePLSDA: Partial least squares-discriminant analysis (PLS-DA)

In HimesGroup/qmtools: Quantitative Metabolomics Data Processing Tools

Partial least squares-discriminant analysis (PLS-DA)

Description

Performs PLS-DA on a matrix-like object where rows represent features and columns represent samples.

Usage

reducePLSDA(x, y, ncomp = 2, center = TRUE, scale = FALSE, ...)

Arguments

x	A matrix-like object.
y	A factor vector for the information about each sample's group.
ncomp	A integer specifying the number of components to extract.
center	A logical specifying whether x and y matrices need to be mean-centered prior to PLS-DA.
scale	A logical specifying whether the unit variance scaling needs to be performed on x prior to PLS-DA.
...	Additional arguments passed to pls::plsr.

Details

This function performs standard PLS for classification with the transpose of x using the pls::plsr. Since PLS-DA is a supervised method, users must supply the information about each sample's group. Here, y must be a factor so that it can be internally converted to an indicator matrix. The function returns a reduced.plsda object that is a matrix with custom attributes to summarize (via summary) and visualize (via plotReduced) the PLS-DA result. The custom attributes include the following:

• method: The method used to reduce the dimension of data.

• ncomp: The number of components extracted.

• explvar: A vector indicating the amount of X variance explained by each component.

• responses: A vector indicating the levels of factor y.

• predictors: A vector of predictor variables.

• coefficient: An array of regression coefficients.

• loadings: A matrix of loadings.

• loadings.weights: A matrix of loading weights.

• Y.observed: A vector of observed responses.

• Y.predicted: A vector of predicted responses.

• Y.scores: A matrix of Y-scores.

• Y.loadings: A matrix of Y-loadings.

• projection: The projection matrix.

• fitted.values: An array of fitted values.

• residuals: An array of regression residuals.

• centered: A logical indicating whether the data was mean-centered prior to PLS-DA.

• scaled: A logical indicating whether the data was scaled prior to PLS-DA.

Value

A reduced.plsda object with the same number of rows as ncol(x) containing the dimension reduction result.

References

Kristian Hovde Liland, Bjørn-Helge Mevik and Ron Wehrens (2021). pls: Partial Least Squares and Principal Component Regression. R package version 2.8-0. https://CRAN.R-project.org/package=pls

See Also

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

See plotReduced for visualization.

See pls::plsr for the underlying function that does the work.

Examples

data(faahko_se)

m <- assay(faahko_se, "knn_vsn")
y <- factor(colData(faahko_se)$sample_group)
res <- reducePLSDA(m, y = y)
summary(res)

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