mixOmics (version 6.2.0)

mint.splsda: P-integration with Discriminant Analysis and variable selection

Description

Function to combine multiple independent studies measured on the same variables or predictors (P-integration) using variants of multi-group sparse PLS-DA for supervised classification with variable selection.

Usage

mint.splsda(X,
Y,
ncomp = 2,
mode = c("regression", "canonical", "invariant", "classic"),
study,
keepX.constraint = NULL,
keepX = rep(ncol(X), ncomp),
scale = TRUE,
tol = 1e-06,
max.iter = 100,
near.zero.var = FALSE)

Arguments

X

numeric matrix of predictors combining multiple independent studies on the same set of predictors. NAs are allowed.

Y

A factor or a class vector indicating the discrete outcome of each sample.

ncomp

Number of components to include in the model (see Details). Default to 2

mode

character string. What type of algorithm to use, (partially) matching one of "regression" or "canonical". See Details.

study

factor indicating the membership of each sample to each of the studies being combined

keepX.constraint

A list of vectors. Each vector entry indicates specific variables to be selected in X for each previous PLS component. See details.

keepX

numeric vector indicating the number of variables to select in X on each component, or remaining components should keepX.constraint be provided. By default all variables are kept in the model.

scale

boleean. If scale = TRUE, each block is standardized to zero means and unit variances. Default = TRUE.

tol

Convergence stopping value.

max.iter

integer, the maximum number of iterations.

near.zero.var

boolean, see the internal nearZeroVar function (should be set to TRUE in particular for data with many zero values). Default = FALSE.

Value

mint.splsda returns an object of class "mint.splsda", "splsda", a list that contains the following components:

X

the centered and standardized original predictor matrix.

Y

the centered and standardized original response vector or matrix.

ind.mat

the centered and standardized original response vector or matrix.

ncomp

the number of components included in the model.

study

The study grouping factor

mode

the algorithm used to fit the model.

keepX

Number of variables used to build each component of X

keepX.constraint

list indicating which variables where constrained to be used to build the first component of X

variates

list containing the variates of X - global variates.

loadings

list containing the estimated loadings for the variates - global loadings.

variates.partial

list containing the variates of X relative to each study - partial variates.

loadings.partial

list containing the estimated loadings for the partial variates - partial loadings.

names

list containing the names to be used for individuals and variables.

nzv

list containing the zero- or near-zero predictors information.

iter

Number of iterations of the algorthm for each component

explained_variance

Percentage of explained variance for each component and each study (note that contrary to PCA, this amount may not decrease as the aim of the method is not to maximise the variance, but the covariance between X and the dummy matrix Y).

Details

mint.splsda function fits a vertical sparse PLS-DA models with ncomp components in which several independent studies measured on the same variables are integrated. The aim is to classify the discrete outcome Y and select variables that explain the outcome. The study factor indicates the membership of each sample in each study. We advise to only combine studies with more than 3 samples as the function performs internal scaling per study, and where all outcome categories are represented.

X can contain missing values. Missing values are handled by being disregarded during the cross product computations in the algorithm mint.splsda without having to delete rows with missing data. Alternatively, missing data can be imputed prior using the nipals function.

The type of algorithm to use is specified with the mode argument. Four PLS algorithms are available: PLS regression ("regression"), PLS canonical analysis ("canonical"), redundancy analysis ("invariant") and the classical PLS algorithm ("classic") (see References and more details in ?pls).

Constraint model. To improve performance, a new argument was recently added to fit a constraint model (constraint=TRUE). The model is run on the optimal list of selected features keepX.constraint from the previous components, as opposed to considering only a specified number of features keepX. Such strategy was implemented in the sister package bootPLS and successfully applied in our recent integrative study (Rohart et al., 2016). Our experience has shown that the constraint model can substantially improve the performance of the method (see tune.mint.splsda).

Variable selection is performed on each component for X via input parameter keepX and keepX.constraint afor a constraint model. keepX.constraint specifies variable names to be selected in the previous components of the model, see example.

Useful graphical outputs are available, e.g. plotIndiv, plotLoadings, plotVar.

References

Rohart F, Eslami A, Matigian, N, Bougeard S, Le Cao K-A (2017). MINT: A multivariate integrative approach to identify a reproducible biomarker signature across multiple experiments and platforms. BMC Bioinformatics 18:128.

Eslami, A., Qannari, E. M., Kohler, A., and Bougeard, S. (2014). Algorithms for multi-group PLS. J. Chemometrics, 28(3), 192-201.

mixOmics manuscript:

Rohart F, Gautier B, Singh A, Le Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. BioRxiv available here: http://biorxiv.org/content/early/2017/02/14/108597

See Also

spls, summary, plotIndiv, plotVar, predict, perf, mint.pls, mint.plsda, mint.plsda and http://www.mixOmics.org/mixMINT for more details.

Examples

# NOT RUN {
data(stemcells)

# -- feature selection with keepX
res = mint.splsda(X = stemcells$gene, Y = stemcells$celltype, ncomp = 3, keepX = c(10, 5, 15),
study = stemcells$study)

plotIndiv(res)
#plot study-specific outputs for all studies
plotIndiv(res, study = "all.partial")

#plot study-specific outputs for study "2"
plotIndiv(res, study = "2")

# -- feature selection with keepX.constraint and keepX
# we force mint.spls to build the PLS-component 1 and 2 with only specific genes
keepX.constraint = list(comp1 = c("ENSG00000138756", "ENSG00000101470"),
comp2 = c("ENSG00000120049", "ENSG00000138685", "ENSG00000204248", "ENSG00000103260"))

# we select 10 and 15 variables on component 3 and 4
res = mint.splsda(X = stemcells$gene, Y = stemcells$celltype, ncomp = 4, study = stemcells$study,
keepX.constraint = keepX.constraint, keepX = c(10, 15))

#plot study-specific outputs for study "2", "3" and "4"
plotIndiv(res, study = c(2, 3, 4))
# poor results as the genes constraint on comp 1 and 2 were not relevant

# }