Package 'mixOmics'

Description

Multivariate methods are well suited to large omics data sets where the number of variables (e.g. genes, proteins, metabolites) is much larger than the number of samples (patients, cells, mice). They have the appealing properties of reducing the dimension of the data by using instrumental variables (components), which are defined as combinations of all variables. Those components are then used to produce useful graphical outputs that enable better understanding of the relationships and correlation structures between the different data sets that are integrated.

Details

mixOmics offers a wide range of multivariate methods for the exploration and integration of biological datasets with a particular focus on variable selection. The package proposes several sparse multivariate models we have developed to identify the key variables that are highly correlated, and/or explain the biological outcome of interest. The data that can be analysed with mixOmics may come from high throughput sequencing technologies, such as omics data (transcriptomics, metabolomics, proteomics, metagenomics etc) but also beyond the realm of omics (e.g. spectral imaging).

The methods implemented in mixOmics can also handle missing values without having to delete entire rows with missing data. A non exhaustive list of methods include variants of generalised Canonical Correlation Analysis, sparse Partial Least Squares and sparse Discriminant Analysis. Recently we implemented integrative methods to combine multiple data sets: N-integration with variants of Generalised Canonical Correlation Analysis and P-integration with variants of multi-group Partial Least Squares.

Description

Calculates the AUC and plots ROC for supervised models from s/plsda, mint.s/plsda and block.plsda, block.splsda or wrapper.sgccda functions.

Usage

auroc(object, ...)

## S3 method for class 'mixo_plsda'
auroc(
  object,
  newdata = object$input.X,
  outcome.test = as.factor(object$Y),
  multilevel = NULL,
  plot = TRUE,
  roc.comp = NULL,
  title = NULL,
  print = TRUE,
  ...
)

## S3 method for class 'mixo_splsda'
auroc(
  object,
  newdata = object$input.X,
  outcome.test = as.factor(object$Y),
  multilevel = NULL,
  plot = TRUE,
  roc.comp = NULL,
  title = NULL,
  print = TRUE,
  ...
)

## S3 method for class 'list'
auroc(object, plot = TRUE, roc.comp = NULL, title = NULL, print = TRUE, ...)

## S3 method for class 'mint.plsda'
auroc(
  object,
  newdata = object$X,
  outcome.test = as.factor(object$Y),
  study.test = object$study,
  multilevel = NULL,
  plot = TRUE,
  roc.comp = NULL,
  roc.study = "global",
  title = NULL,
  print = TRUE,
  ...
)

## S3 method for class 'mint.splsda'
auroc(
  object,
  newdata = object$X,
  outcome.test = as.factor(object$Y),
  study.test = object$study,
  multilevel = NULL,
  plot = TRUE,
  roc.comp = NULL,
  roc.study = "global",
  title = NULL,
  print = TRUE,
  ...
)

## S3 method for class 'sgccda'
auroc(
  object,
  newdata = object$X,
  outcome.test = as.factor(object$Y),
  multilevel = NULL,
  plot = TRUE,
  roc.block = 1L,
  roc.comp = NULL,
  title = NULL,
  print = TRUE,
  ...
)

## S3 method for class 'mint.block.plsda'
auroc(
  object,
  newdata = object$X,
  study.test = object$study,
  outcome.test = as.factor(object$Y),
  multilevel = NULL,
  plot = TRUE,
  roc.block = 1,
  roc.comp = NULL,
  title = NULL,
  print = TRUE,
  ...
)

## S3 method for class 'mint.block.splsda'
auroc(
  object,
  newdata = object$X,
  study.test = object$study,
  outcome.test = as.factor(object$Y),
  multilevel = NULL,
  plot = TRUE,
  roc.block = 1,
  roc.comp = NULL,
  title = NULL,
  print = TRUE,
  ...
)

Arguments

Details

For more than two classes in the categorical outcome Y, the AUC is calculated as one class vs. the other and the ROC curves one class vs. the others are output.

The ROC and AUC are calculated based on the predicted scores obtained from the predict function applied to the multivariate methods (predict(object)$predict). Our multivariate supervised methods already use a prediction threshold based on distances (see predict) that optimally determine class membership of the samples tested. As such AUC and ROC are not needed to estimate the performance of the model (see perf, tune that report classification error rates). We provide those outputs as complementary performance measures.

The pvalue is from a Wilcoxon test between the predicted scores between one class vs the others.

External independent data set (newdata) and outcome (outcome.test) can be input to calculate AUROC. The external data set must have the same variables as the training data set (object$X).

If object is a named list of multiple plsda and splsda objects, ensure that these models each have a response variable with the same levels. Additionally, newdata and outcome.test cannot be passed to this form of auroc.

If newdata is not provided, AUROC is calculated from the training data set, and may result in overfitting (too optimistic results).

Note that for mint.plsda and mint.splsda objects, if roc.study is different from "global", then newdata), outcome.test and sstudy.test are not used.

Value

Depending on the type of object used, a list that contains: The AUC and Wilcoxon test pvalue for each 'one vs other' classes comparison performed, either per component (splsda, plsda, mint.plsda, mint.splsda), or per block and per component (wrapper.sgccda, block.plsda, blocksplsda).

Author(s)

Benoit Gautier, Francois Bartolo, Florian Rohart, Al J Abadi

See Also

tune, perf, and http://www.mixOmics.org for more details.

Examples

## example with PLSDA, 2 classes
# ----------------
data(breast.tumors)
X <- breast.tumors$gene.exp
Y <- breast.tumors$sample$treatment

plsda.breast <- plsda(X, Y, ncomp = 2)
auc.plsda.breast = auroc(plsda.breast, roc.comp = 1)
auc.plsda.breast = auroc(plsda.breast, roc.comp = 2)

## Not run: 
## example with sPLSDA
# -----------------
splsda.breast <- splsda(X, Y, ncomp = 2, keepX = c(25, 25))
auroc(plsda.breast, plot = FALSE)


## example with sPLSDA with 4 classes
# -----------------
data(liver.toxicity)
X <- as.matrix(liver.toxicity$gene)
# Y will be transformed as a factor in the function,
# but we set it as a factor to set up the colors.
Y <- as.factor(liver.toxicity$treatment[, 4])

splsda.liver <- splsda(X, Y, ncomp = 2, keepX = c(20, 20))
auc.splsda.liver = auroc(splsda.liver, roc.comp = 2)


## example with mint.plsda
# -----------------
data(stemcells)

res = mint.plsda(X = stemcells$gene, Y = stemcells$celltype, ncomp = 3,
study = stemcells$study)
auc.mint.pslda = auroc(res, plot = FALSE)

## example with mint.splsda
# -----------------
res = mint.splsda(X = stemcells$gene, Y = stemcells$celltype, ncomp = 3, keepX = c(10, 5, 15),
study = stemcells$study)
auc.mint.spslda = auroc(res, plot = TRUE, roc.comp = 3)


## example with block.plsda
# ------------------
data(nutrimouse)
data = list(gene = nutrimouse$gene, lipid = nutrimouse$lipid)
# with this design, all blocks are connected
design = matrix(c(0,1,1,0), ncol = 2, nrow = 2,
byrow = TRUE, dimnames = list(names(data), names(data)))

block.plsda.nutri = block.plsda(X = data, Y = nutrimouse$diet)
auc.block.plsda.nutri = auroc(block.plsda.nutri, roc.block = 'lipid')

## example with block.splsda
# ---------------
list.keepX = list(gene = rep(10, 2), lipid = rep(5,2))
block.splsda.nutri = block.splsda(X = data, Y = nutrimouse$diet, keepX = list.keepX)
auc.block.splsda.nutri = auroc(block.splsda.nutri, roc.block = 1)

## End(Not run)

Description

Calculate prediction areas that can be used in plotIndiv to shade the background.

Usage

background.predict(
  object,
  comp.predicted = 1,
  dist = "max.dist",
  xlim = NULL,
  ylim = NULL,
  resolution = 100
)

Arguments

Details

background.predict simulates resolution*resolution points within the rectangle defined by xlim on the x-axis and ylim on the y-axis, and then predicts the class of each point (defined by two coordinates). The algorithm estimates the predicted area for each class, defined as the 2D surface where all points are predicted to be of the same class. A polygon is returned and should be passed to plotIndiv for plotting the actual background.

Note that by default xlim and ylim will create a rectangle of simulated data that will cover the plotted area of plotIndiv. However, if you use plotIndiv with ellipse=TRUE or if you set xlim and ylim, then you will need to adapt xlim and ylim in background.predict.

Also note that the white frontier that defines the predicted areas when plotting with plotIndiv can be reduced by increasing resolution.

More details about the prediction distances in ?predict and the supplemental material of the mixOmics article (Rohart et al. 2017).

Value

background.predict returns a list of coordinates to be used with polygon to draw the predicted area for each class.

Author(s)

Florian Rohart, Al J Abadi

References

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

See Also

plotIndiv, predict, polygon.

Examples

# Example 1
# -----------------------------------
data(breast.tumors)
X <- breast.tumors$gene.exp
Y <- breast.tumors$sample$treatment

splsda.breast <- splsda(X, Y,keepX=c(10,10),ncomp=2)

# calculating background for the two first components, and the centroids distance

background = background.predict(splsda.breast, comp.predicted = 2, dist = "centroids.dist")

## Not run: 
# default option: note that the outcome color is included by default!
plotIndiv(splsda.breast, background = background, legend=TRUE)

# Example 2
# -----------------------------------
data(liver.toxicity)
X = liver.toxicity$gene
Y = as.factor(liver.toxicity$treatment[, 4])

plsda.liver <- plsda(X, Y, ncomp = 2)

# calculating background for the two first components, and the mahalanobis distance
background = background.predict(plsda.liver, comp.predicted = 2, dist = "mahalanobis.dist")

plotIndiv(plsda.liver, background = background, legend = TRUE)



## End(Not run)

Description

biplot methods for pca family

Usage

## S3 method for class 'pca'
biplot(
  x,
  comp = c(1, 2),
  block = NULL,
  ind.names = TRUE,
  group = NULL,
  cutoff = 0,
  col.per.group = NULL,
  col = NULL,
  ind.names.size = 3,
  ind.names.col = color.mixo(4),
  ind.names.repel = TRUE,
  pch = 19,
  pch.levels = NULL,
  pch.size = 2,
  var.names = TRUE,
  var.names.col = "grey40",
  var.names.size = 4,
  var.names.angle = FALSE,
  var.arrow.col = "grey40",
  var.arrow.size = 0.5,
  var.arrow.length = 0.2,
  ind.legend.title = NULL,
  vline = FALSE,
  hline = FALSE,
  legend = if (is.null(group)) FALSE else TRUE,
  legend.title = NULL,
  pch.legend.title = NULL,
  cex = 1.05,
  ...
)

## S3 method for class 'mixo_pls'
biplot(
  x,
  comp = c(1, 2),
  block = NULL,
  ind.names = TRUE,
  group = NULL,
  cutoff = 0,
  col.per.group = NULL,
  col = NULL,
  ind.names.size = 3,
  ind.names.col = color.mixo(4),
  ind.names.repel = TRUE,
  pch = 19,
  pch.levels = NULL,
  pch.size = 2,
  var.names = TRUE,
  var.names.col = "grey40",
  var.names.size = 4,
  var.names.angle = FALSE,
  var.arrow.col = "grey40",
  var.arrow.size = 0.5,
  var.arrow.length = 0.2,
  ind.legend.title = NULL,
  vline = FALSE,
  hline = FALSE,
  legend = if (is.null(group)) FALSE else TRUE,
  legend.title = NULL,
  pch.legend.title = NULL,
  cex = 1.05,
  ...
)

Arguments

Details

biplot unifies the reduced representation of both the observations/samples and variables of a matrix of multivariate data on the same plot. Essentially, in the reduced space the samples are shown as points/names and the contributions of features to each dimension are shown as directed arrows or vectors. For pls objects it is possible to use either 'X' or 'Y' latent space using block argument.

Value

A ggplot object.

Author(s)

Al J Abadi

Examples

data("nutrimouse")
## --------- pca ---------- ##
pca.lipid <- pca(nutrimouse$lipid, ncomp = 3, scale = TRUE)
# seed for reproducible geom_text_repel
set.seed(42)
biplot(pca.lipid)
## correlation cutoff to filter features
biplot(pca.lipid, cutoff = c(0.8))
## tailor threshold for each component
biplot(pca.lipid, cutoff = c(0.8, 0.7))
## customise components
biplot(pca.lipid, cutoff = c(0.8), comp = c(1,3))

## customise ggplot in an arbitrary way
biplot(pca.lipid) + theme_linedraw() + 
    # add vline
    geom_vline(xintercept = 0, col = 'green') +
    # add hline
    geom_hline(yintercept = 0, col = 'green') +
    # customise labs
    labs(x = 'Principal Component 1', y = 'Principal Component 2')

## group samples
biplot(pca.lipid, group = nutrimouse$diet, legend.title = 'Diet')

## customise variable labels
biplot(pca.lipid, 
       var.names.col = color.mixo(2),
       var.names.size = 4,
       var.names.angle = TRUE
)

## no arrows
biplot(pca.lipid, group = nutrimouse$diet, legend.title = 'Diet', 
       var.arrow.col = NULL, var.names.col = 'black')

## add x=0 and y=0 lines in function
biplot(pca.lipid, group = nutrimouse$diet, legend.title = 'Diet', 
       var.arrow.col = NULL, var.names.col = 'black', 
       vline = TRUE, hline = TRUE)

## --------- spca
## example with spca
spca.lipid <- spca(nutrimouse$lipid, ncomp = 2, scale = TRUE, keepX = c(8, 6))
biplot(spca.lipid, var.names.col = 'black', group = nutrimouse$diet, 
       legend.title = 'Diet')

## --------- pls ---------- ##
data("nutrimouse")
pls.nutrimouse <- pls(X = nutrimouse$gene, Y = nutrimouse$lipid, ncomp = 2)
biplot(pls.nutrimouse, group = nutrimouse$genotype, block = 'X',
       legend.title = 'Genotype', cutoff = 0.878) 

biplot(pls.nutrimouse, group = nutrimouse$genotype, block = 'Y',
       legend.title = 'Genotype', cutoff = 0.8) 

## --------- plsda ---------- ##
data(breast.tumors)
X <- breast.tumors$gene.exp
colnames(X) <- paste0('GENE_', colnames(X))
rownames(X) <- paste0('SAMPLE_', rownames(X))
Y <- breast.tumors$sample$treatment

plsda.breast <- plsda(X, Y, ncomp = 2)
biplot(plsda.breast, cutoff = 0.72)
## remove arrows
biplot(plsda.breast, cutoff = 0.72, var.arrow.col = NULL, var.names.size = 4)

Description

Integration of multiple data sets measured on the same samples or observations, ie. N-integration. The method is partly based on Generalised Canonical Correlation Analysis.

Usage

block.pls(
  X,
  Y,
  indY,
  ncomp = 2,
  design,
  scheme,
  mode,
  scale = TRUE,
  init,
  tol = 1e-06,
  max.iter = 100,
  near.zero.var = FALSE,
  all.outputs = TRUE,
  verbose.call = FALSE
)

Arguments

Details

block.spls function fits a horizontal integration PLS model with a specified number of components per block). An outcome needs to be provided, either by Y or by its position indY in the list of blocks X. Multi (continuous)response are supported. X and Y can contain missing values. Missing values are handled by being disregarded during the cross product computations in the algorithm block.pls without having to delete rows with missing data. Alternatively, missing data can be imputed prior using the impute.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 ?pls for more details).

Note that our method is partly based on Generalised Canonical Correlation Analysis and differs from the MB-PLS approaches proposed by Kowalski et al., 1989, J Chemom 3(1) and Westerhuis et al., 1998, J Chemom, 12(5).

Value

block.pls returns an object of class 'block.pls', a list that contains the following components:

Author(s)

Florian Rohart, Benoit Gautier, Kim-Anh Lê Cao, Al J Abadi

References

Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions Technic.

Wold H. (1966). Estimation of principal components and related models by iterative least squares. In: Krishnaiah, P. R. (editors), Multivariate Analysis. Academic Press, N.Y., 391-420.

Tenenhaus A. and Tenenhaus M., (2011), Regularized Generalized Canonical Correlation Analysis, Psychometrika, Vol. 76, Nr 2, pp 257-284.

See Also

plotIndiv, plotArrow, plotLoadings, plotVar, predict, perf, selectVar, block.spls, block.plsda and http://www.mixOmics.org for more details.

Examples

# Example with TCGA multi omics study
# -----------------------------------
data("breast.TCGA")
# this is the X data as a list of mRNA and miRNA; the Y data set is a single data set of proteins
data = list(mrna = breast.TCGA$data.train$mrna, mirna = breast.TCGA$data.train$mirna)
# set up a full design where every block is connected
design = matrix(1, ncol = length(data), nrow = length(data),
dimnames = list(names(data), names(data)))
diag(design) =  0
design
# set number of component per data set
ncomp = c(2)

TCGA.block.pls = block.pls(X = data, Y = breast.TCGA$data.train$protein, 
                           ncomp = ncomp, design = design)
TCGA.block.pls
## use design = 'full'
TCGA.block.pls = block.pls(X = data, Y = breast.TCGA$data.train$protein, 
                           ncomp = ncomp, design = 'full')
# in plotindiv we color the samples per breast subtype group but the method is unsupervised!
# here Y is the protein data set
plotIndiv(TCGA.block.pls, group =  breast.TCGA$data.train$subtype, ind.names = FALSE)

Description

Integration of multiple data sets measured on the same samples or observations to classify a discrete outcome, ie. N-integration with Discriminant Analysis. The method is partly based on Generalised Canonical Correlation Analysis.

Usage

block.plsda(
  X,
  Y,
  indY,
  ncomp = 2,
  design,
  scheme,
  scale = TRUE,
  init = "svd",
  tol = 1e-06,
  max.iter = 100,
  near.zero.var = FALSE,
  all.outputs = TRUE,
  verbose.call = FALSE
)

Arguments

Details

block.plsda function fits a horizontal integration PLS-DA model with a specified number of components per block). A factor indicating the discrete outcome needs to be provided, either by Y or by its position indY in the list of blocks X.

X can contain missing values. Missing values are handled by being disregarded during the cross product computations in the algorithm block.pls without having to delete rows with missing data. Alternatively, missing data can be imputed prior using the impute.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 ?pls for more details).

Note that our method is partly based on Generalised Canonical Correlation Analysis and differs from the MB-PLS approaches proposed by Kowalski et al., 1989, J Chemom 3(1) and Westerhuis et al., 1998, J Chemom, 12(5).

Value

block.plsda returns an object of class "block.plsda","block.pls", a list that contains the following components:

Author(s)

Florian Rohart, Benoit Gautier, Kim-Anh Lê Cao, Al J Abadi

References

On PLSDA:

Barker M and Rayens W (2003). Partial least squares for discrimination. Journal of Chemometrics 17(3), 166-173. Perez-Enciso, M. and Tenenhaus, M. (2003). Prediction of clinical outcome with microarray data: a partial least squares discriminant analysis (PLS-DA) approach. Human Genetics 112, 581-592. Nguyen, D. V. and Rocke, D. M. (2002). Tumor classification by partial least squares using microarray gene expression data. Bioinformatics 18, 39-50.

On multiple integration with PLS-DA: Gunther O., Shin H., Ng R. T. , McMaster W. R., McManus B. M. , Keown P. A. , Tebbutt S.J. , Lê Cao K-A. , (2014) Novel multivariate methods for integration of genomics and proteomics data: Applications in a kidney transplant rejection study, OMICS: A journal of integrative biology, 18(11), 682-95.

On multiple integration with sPLS-DA and 4 data blocks:

Singh A., Gautier B., Shannon C., Vacher M., Rohart F., Tebbutt S. and Lê Cao K.A. (2016). DIABLO: multi omics integration for biomarker discovery. BioRxiv available here: http://biorxiv.org/content/early/2016/08/03/067611

mixOmics article:

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

See Also

plotIndiv, plotArrow, plotLoadings, plotVar, predict, perf, selectVar, block.pls, block.splsda and http://www.mixOmics.org for more details.

Examples

data(nutrimouse)
data = list(gene = nutrimouse$gene, lipid = nutrimouse$lipid, Y = nutrimouse$diet)
# with this design, all blocks are connected
design = matrix(c(0,1,1,1,0,1,1,1,0), ncol = 3, nrow = 3,
byrow = TRUE, dimnames = list(names(data), names(data)))

res = block.plsda(X = data, indY = 3) # indY indicates where the outcome Y is in the list X
plotIndiv(res, ind.names = FALSE, legend = TRUE)
plotVar(res)

## Not run: 
# when Y is provided
res2 = block.plsda(list(gene = nutrimouse$gene, lipid = nutrimouse$lipid),
Y = nutrimouse$diet, ncomp = 2)
plotIndiv(res2)
plotVar(res2)

## End(Not run)

Description

Integration of multiple data sets measured on the same samples or observations, with variable selection in each data set, ie. N-integration. The method is partly based on Generalised Canonical Correlation Analysis.

Usage

block.spls(
  X,
  Y,
  indY,
  ncomp = 2,
  keepX,
  keepY,
  design,
  scheme,
  mode,
  scale = TRUE,
  init,
  tol = 1e-06,
  max.iter = 100,
  near.zero.var = FALSE,
  all.outputs = TRUE,
  verbose.call = FALSE
)

Arguments

Details

block.spls function fits a horizontal sPLS model with a specified number of components per block). An outcome needs to be provided, either by Y or by its position indY in the list of blocks X. Multi (continuous)response are supported. X and Y can contain missing values. Missing values are handled by being disregarded during the cross product computations in the algorithm block.pls 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 ?pls for more details).

Note that our method is partly based on sparse Generalised Canonical Correlation Analysis and differs from the MB-PLS approaches proposed by Kowalski et al., 1989, J Chemom 3(1), Westerhuis et al., 1998, J Chemom, 12(5) and sparse variants Li et al., 2012, Bioinformatics 28(19); Karaman et al (2014), Metabolomics, 11(2); Kawaguchi et al., 2017, Biostatistics.

Variable selection is performed on each component for each block of X, and for Y if specified, via input parameter keepX and keepY.

Note that if Y is missing and indY is provided, then variable selection on Y is performed by specifying the input parameter directly in keepX (no keepY is needed).

Value

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

Author(s)

Florian Rohart, Benoit Gautier, Kim-Anh Lê Cao, Al J Abadi

References

Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions Technic.

Wold H. (1966). Estimation of principal components and related models by iterative least squares. In: Krishnaiah, P. R. (editors), Multivariate Analysis. Academic Press, N.Y., 391-420.

Tenenhaus A. and Tenenhaus M., (2011), Regularized Generalized Canonical Correlation Analysis, Psychometrika, Vol. 76, Nr 2, pp 257-284.

Tenenhaus A., Philippe C., Guillemot V, Lê Cao K.A., Grill J, Frouin V. Variable selection for generalized canonical correlation analysis. Biostatistics. kxu001

See Also

plotIndiv, plotArrow, plotLoadings, plotVar, predict, perf, selectVar, block.pls, block.splsda and http://www.mixOmics.org for more details.

Examples

# Example with multi omics TCGA study
# -----------------------------
data("breast.TCGA")
# this is the X data as a list of mRNA and miRNA; the Y data set is a single data set of proteins
data = list(mrna = breast.TCGA$data.train$mrna, mirna = breast.TCGA$data.train$mirna)
# set up a full design where every block is connected
design = matrix(1, ncol = length(data), nrow = length(data),
dimnames = list(names(data), names(data)))
diag(design) =  0
design
# set number of component per data set
ncomp = c(2)
# set number of variables to select, per component and per data set (this is set arbitrarily)
list.keepX = list(mrna = rep(10, 2), mirna = rep(10,2))
list.keepY = c(rep(10, 2))

TCGA.block.spls = block.spls(X = data, Y = breast.TCGA$data.train$protein,
ncomp = ncomp, keepX = list.keepX, keepY = list.keepY, design = design)
TCGA.block.spls
# in plotindiv we color the samples per breast subtype group but the method is unsupervised!
plotIndiv(TCGA.block.spls, group =  breast.TCGA$data.train$subtype, ind.names = FALSE, legend=TRUE)
# illustrates coefficient weights in each block
plotLoadings(TCGA.block.spls, ncomp = 1)
plotVar(TCGA.block.spls, style = 'graphics', legend = TRUE)

## plot markers (selected markers) for mrna and mirna
group <- breast.TCGA$data.train$subtype
# mrna: show each selected feature separately and group by subtype
plotMarkers(object = TCGA.block.spls, comp = 1, block = 'mrna', group = group)
# mrna: aggregate all selected features, separate by loadings signs and group by subtype
plotMarkers(object = TCGA.block.spls, comp = 1, block = 'mrna', group = group, global = TRUE)
# proteins
plotMarkers(object = TCGA.block.spls, comp = 1, block = 'Y', group = group)
## only show boxplots
plotMarkers(object = TCGA.block.spls, comp = 1, block = 'Y', group = group, violin = FALSE)

## Not run: 
network(TCGA.block.spls)

## End(Not run)

Description

Integration of multiple data sets measured on the same samples or observations to classify a discrete outcome to classify a discrete outcome and select features from each data set, ie. N-integration with sparse Discriminant Analysis. The method is partly based on Generalised Canonical Correlation Analysis.

Usage

block.splsda(
  X,
  Y,
  indY,
  ncomp = 2,
  keepX,
  design,
  scheme,
  scale = TRUE,
  init = "svd",
  tol = 1e-06,
  max.iter = 100,
  near.zero.var = FALSE,
  all.outputs = TRUE,
  verbose.call = FALSE
)

wrapper.sgccda(
  X,
  Y,
  indY,
  ncomp = 2,
  keepX,
  design,
  scheme,
  scale = TRUE,
  init = "svd",
  tol = 1e-06,
  max.iter = 100,
  near.zero.var = FALSE,
  all.outputs = TRUE,
  verbose.call = FALSE
)

Arguments

Details

block.splsda function fits a horizontal integration PLS-DA model with a specified number of components per block). A factor indicating the discrete outcome needs to be provided, either by Y or by its position indY in the list of blocks X.

X can contain missing values. Missing values are handled by being disregarded during the cross product computations in the algorithm block.pls without having to delete rows with missing data. Alternatively, missing data can be imputed prior using the impute.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 ?pls for more details).

Note that our method is partly based on sparse Generalised Canonical Correlation Analysis and differs from the MB-PLS approaches proposed by Kowalski et al., 1989, J Chemom 3(1), Westerhuis et al., 1998, J Chemom, 12(5) and sparse variants Li et al., 2012, Bioinformatics 28(19); Karaman et al (2014), Metabolomics, 11(2); Kawaguchi et al., 2017, Biostatistics.

Variable selection is performed on each component for each block of X if specified, via input parameter keepX.

Value

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

Author(s)

Florian Rohart, Benoit Gautier, Kim-Anh Lê Cao, Al J Abadi

References

On multiple integration with sPLS-DA and 4 data blocks:

Singh A., Gautier B., Shannon C., Vacher M., Rohart F., Tebbutt S. and Lê Cao K.A. (2016). DIABLO: multi omics integration for biomarker discovery. BioRxiv available here: http://biorxiv.org/content/early/2016/08/03/067611

On data integration:

Tenenhaus A., Philippe C., Guillemot V, Lê Cao K.A., Grill J, Frouin V. Variable selection for generalized canonical correlation analysis. Biostatistics. kxu001

Gunther O., Shin H., Ng R. T. , McMaster W. R., McManus B. M. , Keown P. A. , Tebbutt S.J. , Lê Cao K-A. , (2014) Novel multivariate methods for integration of genomics and proteomics data: Applications in a kidney transplant rejection study, OMICS: A journal of integrative biology, 18(11), 682-95.

mixOmics article:

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

See Also

plotIndiv, plotArrow, plotLoadings, plotVar, predict, perf, selectVar, block.plsda, block.spls and http://www.mixOmics.org/mixDIABLO for more details and examples.

Examples

# block.splsda
# -------------
data("breast.TCGA")
# this is the X data as a list of mRNA, miRNA and proteins
data = list(mrna = breast.TCGA$data.train$mrna, mirna = breast.TCGA$data.train$mirna,
protein = breast.TCGA$data.train$protein)
# set up a full design where every block is connected
design = matrix(1, ncol = length(data), nrow = length(data),
dimnames = list(names(data), names(data)))
diag(design) =  0
design
# set number of component per data set
ncomp = c(2)
# set number of variables to select, per component and per data set (this is set arbitrarily)
list.keepX = list(mrna = rep(8,2), mirna = rep(8,2), protein = rep(8,2))


TCGA.block.splsda = block.splsda(X = data, Y = breast.TCGA$data.train$subtype, 
                                 ncomp = ncomp, keepX = list.keepX, design = design)
## use design = 'full'
TCGA.block.splsda = block.splsda(X = data, Y = breast.TCGA$data.train$subtype, 
                                 ncomp = ncomp, keepX = list.keepX, design = 'full')
TCGA.block.splsda$design

plotIndiv(TCGA.block.splsda, ind.names = FALSE)
## use design = 'null'
TCGA.block.splsda = block.splsda(X = data, Y = breast.TCGA$data.train$subtype, 
                                 ncomp = ncomp, keepX = list.keepX, design = 'null')
TCGA.block.splsda$design
## set all off-diagonal elements to 0.5
TCGA.block.splsda = block.splsda(X = data, Y = breast.TCGA$data.train$subtype, 
                                 ncomp = ncomp, keepX = list.keepX, design = 0.5)
TCGA.block.splsda$design
# illustrates coefficient weights in each block
plotLoadings(TCGA.block.splsda, ncomp = 1, contrib = 'max')
plotVar(TCGA.block.splsda, style = 'graphics', legend = TRUE)

## plot markers (selected variables) for mrna and mirna
# mrna: show each selected feature separately
plotMarkers(object = TCGA.block.splsda, comp = 1, block = 'mrna')
# mrna: aggregate all selected features and separate by loadings signs
plotMarkers(object = TCGA.block.splsda, comp = 1, block = 'mrna', global = TRUE)
# proteins
plotMarkers(object = TCGA.block.splsda, comp = 1, block = 'protein')
## do not show violin plots
plotMarkers(object = TCGA.block.splsda, comp = 1, block = 'protein', violin = FALSE)
# show top 5 markers
plotMarkers(object = TCGA.block.splsda, comp = 1, block = 'protein', markers = 1:5)
# show specific markers
my.markers <- selectVar(TCGA.block.splsda, comp = 1)[['protein']]$name[c(1,3,5)]
my.markers
plotMarkers(object = TCGA.block.splsda, comp = 1, block = 'protein', markers = my.markers)

Description

This data set is a small subset of the full data set from The Cancer Genome Atlas that can be analysed with the DIABLO framework. It contains the expression or abundance of three matching omics data sets: mRNA, miRNA and proteomics for 150 breast cancer samples (Basal, Her2, Luminal A) in the training set, and 70 samples in the test set. The test set is missing the proteomics data set.

Usage

data(breast.TCGA)

Format

A list containing two data sets, data.train and data.test which both include:

list("miRNA")

data frame with 150 (70) rows and 184 columns in the training (test) data set. The expression levels of 184 miRNA.

list("mRNA")

data frame with 150 (70) rows and 520 columns in the training (test) data set. The expression levels of 200 mRNA.

list("protein")

data frame with 150 (70) rows and 142 columns in the training data set only. The abundance of 142 proteins.

list("subtype")

a factor indicating the brerast cancer subtypes in the training (length of 150) and test (length of 70) sets.

Details

The data come from The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov/). We divided the data into a training (discovery) and test (validation) set. The protein dataset which had a limited number of subjects available was used to allocate subjects into the training set only, while the tes set included all remaining subject. Each data set was normalised and pre-processed. For illustrative purposes we drastically filtered the data here.

Value

none

Source

The raw data were downloaded from http://cancergenome.nih.gov/. The normalised and filtered data we analysed with DIABLO are available on www.mixOmics.org/mixDIABLO

References

Singh A., Shannon C., Gautier B., Rohart F., Vacher M., Tebbutt S. and Lê Cao K.A. (2019), DIABLO: an integrative approach for identifying key molecular drivers from multi-omics assays, Bioinformatics, Volume 35, Issue 17, 1 September 2019, Pages 3055–3062.

Description

This data set contains the expression of 1,000 genes in 47 surgical specimens of human breast tumours from 17 different individuals before and after chemotherapy treatment.

Usage

data(breast.tumors)

Format

A list containing the following components:

list("gene.exp")

data matrix with 47 rows and 1000 columns. Each row represents an experimental sample, and each column a single gene.

list("sample")

a list containing two character vector components: name the name of the samples, and treatment the treatment status.

list("genes")

a list containing two character vector components: name the name of the genes, and description the description of each gene.

Details

This data consists of 47 breast cancer samples and 1753 cDNA clones pre-selected by Perez-Enciso et al. (2003) to draw their Fig. 1. The authors selected 47 samples for which there was information at least before or before and after chemotherapy treatment. There were 20 tumours that were microarrayed both before and after treatment. For illustrative purposes we then randomly selected 1000 cDNA clones for this data set.

Value

none

Source

The Human Breast Tumors dataset is a companion resource for the paper of Perou et al. (2000), and was downloaded from the Stanford Genomics Breast Cancer Consortium Portal http://genome-www.stanford.edu/breast_cancer/molecularportraits/download.shtml

References

Perez-Enciso, M. and Tenenhaus, M. (2003). Prediction of clinical outcome with microarray data: a partial least squares discriminant analysis (PLS-DA) approach. Human Genetics 112, 581-592.

Perou, C. M., Sorlie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S., Rees, C. A., Pollack, J. R., Ross, D. T., Johnsen, H., Akslen, L. A., Fluge, O., Pergamenschikov, A., Williams, C., Zhu, S. X., Lonning, P. E., Borresen-Dale, A. L., Brown, P. O. and Botstein, D. (2000). Molecular portraits of human breast tumours. Nature 406, 747-752.

Description

This function generates color-coded Clustered Image Maps (CIMs) ("heat maps") to represent "high-dimensional" data sets.

Usage

cim(
  mat = NULL,
  color = NULL,
  row.names = TRUE,
  col.names = TRUE,
  row.sideColors = NULL,
  col.sideColors = NULL,
  row.cex = NULL,
  col.cex = NULL,
  cutoff = 0,
  cluster = "both",
  dist.method = c("euclidean", "euclidean"),
  clust.method = c("complete", "complete"),
  cut.tree = c(0, 0),
  transpose = FALSE,
  symkey = TRUE,
  keysize = c(1, 1),
  keysize.label = 1,
  zoom = FALSE,
  title = NULL,
  xlab = NULL,
  ylab = NULL,
  margins = c(5, 5),
  lhei = NULL,
  lwid = NULL,
  comp = NULL,
  center = TRUE,
  scale = FALSE,
  mapping = "XY",
  legend = NULL,
  save = NULL,
  name.save = NULL,
  blocks = NULL
)

Arguments

Details

One matrix Clustered Image Map (default method) is a 2-dimensional visualization of a real-valued matrix (basically image(t(mat))) with rows and/or columns reordered according to some hierarchical clustering method to identify interesting patterns. Generated dendrograms from clustering are added to the left side and to the top of the image. By default the used clustering method for rows and columns is the complete linkage method and the used distance measure is the distance euclidean.

In "pca", "spca", "ipca", "sipca", "plsda", "splsda" and multilevel variants methods the mat matrix is object$X.

For the remaining methods, if mapping = "X" or mapping = "Y" the mat matrix is object$X or object$Y respectively. If mapping = "XY":

• in rcc method, the matrix mat is created where element $(j,k)$ is the scalar product value between every pairs of vectors in dimension length(comp) representing the variables $X_j$ and $Y_k$ on the axis defined by $Z_i$ with $i$ in comp, where $Z_i$ is the equiangular vector between the $i$-th $X$ and $Y$ canonical variate.

• in pls, spls and multilevel spls methods, if object$mode is "regression", the element $(j,k)$ of the matrix mat is given by the scalar product value between every pairs of vectors in dimension length(comp) representing the variables $X_j$ and $Y_k$ on the axis defined by $U_i$ with $i$ in comp, where $U_i$ is the $i$-th $X$ variate. If object$mode is "canonical" then $X_j$ and $Y_k$ are represented on the axis defined by $U_i$ and $V_i$ respectively.

The blocks parameter controls which blocks are to be included when class(mat) == "block.pls" OR "block.spls". This can be a character or a integer vector.

If using a multiblock object then mapping can be set to "multiblock". When done so, this will emulate the function of cimDiablo(), such that rows will denote each sample and all features included in blocks will be shown as columns, coloured by which block they inherit from. In this case, blocks can include any number of input blocks. If mapping = "X", "Y" OR "XY", then it functions similarly to if a mixo_pls object was being used. blocks has to be of length 2 in this scenario.

By default four components will be displayed in the plot. At the top left is the color key, top right is the column dendogram, bottom left is the row dendogram, bottom right is the image plot. When sideColors are provided, an additional row or column is inserted in the appropriate location. This layout can be overriden by specifiying appropriate values for lwid and lhei. lwid controls the column width, and lhei controls the row height. See the help page for layout for details on how to use these arguments.

For visualization of "high-dimensional" data sets, a nice zooming tool was created. zoom = TRUE open a new device, one for CIM, one for zoom-out region and define an interactive 'zoom' process: click two points at imagen map region by pressing the first mouse button. It then draws a rectangle around the selected region and zoom-out this at new device. The process can be repeated to zoom-out other regions of interest.

The zoom process is terminated by clicking the second button and selecting 'Stop' from the menu, or from the 'Stop' menu on the graphics window.

Value

A list containing the following components:

Author(s)

Ignacio González, Francois Bartolo, Kim-Anh Lê Cao, Al J Abadi

References

Eisen, M. B., Spellman, P. T., Brown, P. O. and Botstein, D. (1998). Cluster analysis and display of genome-wide expression patterns. Proceeding of the National Academy of Sciences of the USA 95, 14863-14868.

Weinstein, J. N., Myers, T. G., O'Connor, P. M., Friend, S. H., Fornace Jr., A. J., Kohn, K. W., Fojo, T., Bates, S. E., Rubinstein, L. V., Anderson, N. L., Buolamwini, J. K., van Osdol, W. W., Monks, A. P., Scudiero, D. A., Sausville, E. A., Zaharevitz, D. W., Bunow, B., Viswanadhan, V. N., Johnson, G. S., Wittes, R. E. and Paull, K. D. (1997). An information-intensive approach to the molecular pharmacology of cancer. Science 275, 343-349.

González I., Lê Cao K.A., Davis M.J., Déjean S. (2012). Visualising associations between paired 'omics' data sets. BioData Mining; 5(1).

mixOmics article:

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

See Also

heatmap, hclust, plotVar, network and

http://mixomics.org/graphics/ for more details on all options available.

Examples

## default method: shows cross correlation between 2 data sets
#------------------------------------------------------------------
data(nutrimouse)
X <- nutrimouse$lipid
Y <- nutrimouse$gene

cim(cor(X, Y), cluster = "none")


## Not run: 
## CIM representation for objects of class 'rcc'
#------------------------------------------------------------------

nutri.rcc <- rcc(X, Y, ncomp = 3, lambda1 = 0.064, lambda2 = 0.008)

cim(nutri.rcc, xlab = "genes", ylab = "lipids", margins = c(5, 6))

#-- interactive 'zoom' available as below

cim(nutri.rcc, xlab = "genes", ylab = "lipids", margins = c(5, 6),
zoom = TRUE)
#-- select the region and "see" the zoom-out region


#-- cim from X matrix with a side bar to indicate the diet
diet.col <- palette()[as.numeric(nutrimouse$diet)]
cim(nutri.rcc, mapping = "X", row.names = nutrimouse$diet,
row.sideColors = diet.col, xlab = "lipids",
clust.method = c("ward", "ward"), margins = c(6, 4))

#-- cim from Y matrix with a side bar to indicate the genotype
geno.col = color.mixo(as.numeric(nutrimouse$genotype))
cim(nutri.rcc, mapping = "Y", row.names = nutrimouse$genotype,
row.sideColors = geno.col, xlab = "genes",
clust.method = c("ward", "ward"))

#-- save the result as a jpeg file
jpeg(filename = "test.jpeg", res = 600, width = 4000, height = 4000)
cim(nutri.rcc, xlab = "genes", ylab = "lipids", margins = c(5, 6))
dev.off()

## CIM representation for objects of class 'spca' (also works for sipca)
#------------------------------------------------------------------

data(liver.toxicity)
X <- liver.toxicity$gene

liver.spca <- spca(X, ncomp = 2, keepX = c(30, 30), scale = FALSE)

dose.col <- color.mixo(as.numeric(as.factor(liver.toxicity$treatment[, 3])))

# side bar, no variable names shown
cim(liver.spca, row.sideColors = dose.col, col.names = FALSE,
row.names = liver.toxicity$treatment[, 3],
clust.method = c("ward", "ward"))


## CIM representation for objects of class '(s)pls'
#------------------------------------------------------------------

data(liver.toxicity)

X <- liver.toxicity$gene
Y <- liver.toxicity$clinic
liver.spls <- spls(X, Y, ncomp = 3,
keepX = c(20, 50, 50), keepY = c(10, 10, 10))


# default
cim(liver.spls)


# transpose matrix, choose clustering method
cim(liver.spls, transpose = TRUE,
clust.method = c("ward", "ward"), margins = c(5, 7))

# Here we visualise only the X variables selected
cim(liver.spls, mapping="X")

# Here we should visualise only the Y variables selected
cim(liver.spls, mapping="Y")

# Here we only visualise the similarity matrix between the variables by spls
cim(liver.spls, cluster="none")

# plotting two data sets with the similarity matrix as input in the funciton
# (see our BioData Mining paper for more details)
# Only the variables selected by the sPLS model in X and Y are represented
cim(liver.spls, mapping="XY")

# on the X matrix only, side col var to indicate dose
dose.col <- color.mixo(as.numeric(as.factor(liver.toxicity$treatment[, 3])))
cim(liver.spls, mapping = "X", row.sideColors = dose.col,
row.names = liver.toxicity$treatment[, 3])

# CIM default representation includes the total of 120 genes selected, with the dose color
# with a sparse method, show only the variables selected on specific components
cim(liver.spls, comp = 1)
cim(liver.spls, comp = 2)
cim(liver.spls, comp = c(1,2))
cim(liver.spls, comp = c(1,3))


## CIM representation for objects of class '(s)plsda'
#------------------------------------------------------------------
data(liver.toxicity)

X <- liver.toxicity$gene
# Setting up the Y outcome first
Y <- liver.toxicity$treatment[, 3]
#set up colors for cim
dose.col <- color.mixo(as.numeric(as.factor(liver.toxicity$treatment[, 3])))


liver.splsda <- splsda(X, Y, ncomp = 2, keepX = c(40, 30))

cim(liver.splsda, row.sideColors = dose.col, row.names = Y)


## CIM representation for objects of class splsda 'multilevel'
# with a two level factor (repeated sample and time)
#------------------------------------------------------------------
data(vac18.simulated)
X <- vac18.simulated$genes
design <- data.frame(samp = vac18.simulated$sample)
Y = data.frame(time = vac18.simulated$time,
stim = vac18.simulated$stimulation)

res.2level <- splsda(X, Y = Y, ncomp = 2, multilevel = design,
keepX = c(120, 10))

#define colors for the levels: stimulation and time
stim.col <- c("darkblue", "purple", "green4","red3")
stim.col <- stim.col[as.numeric(Y$stim)]
time.col <- c("orange", "cyan")[as.numeric(Y$time)]


# The row side bar indicates the two levels of the facteor, stimulation and time.
# the sample names have been motified on the plot.
cim(res.2level, row.sideColors = cbind(stim.col, time.col),
row.names = paste(Y$time, Y$stim, sep = "_"),
col.names = FALSE,
#setting up legend:
legend=list(legend = c(levels(Y$time), levels(Y$stim)),
col = c("orange", "cyan", "darkblue", "purple", "green4","red3"),
title = "Condition", cex = 0.7)
)

## CIM representation for objects of class spls 'multilevel'
#------------------------------------------------------------------

data(liver.toxicity)
repeat.indiv <- c(1, 2, 1, 2, 1, 2, 1, 2, 3, 3, 4, 3, 4, 3, 4, 4, 5, 6, 5, 5,
6, 5, 6, 7, 7, 8, 6, 7, 8, 7, 8, 8, 9, 10, 9, 10, 11, 9, 9,
10, 11, 12, 12, 10, 11, 12, 11, 12, 13, 14, 13, 14, 13, 14,
13, 14, 15, 16, 15, 16, 15, 16, 15, 16)

# sPLS is a non supervised technique, and so we only indicate the sample repetitions
# in the design (1 factor only here, sample)
# sPLS takes as an input 2 data sets, and the variables selected
design <- data.frame(sample = repeat.indiv)
res.spls.1level <- spls(X = liver.toxicity$gene,
Y=liver.toxicity$clinic,
multilevel = design,
ncomp = 2,
keepX = c(50, 50), keepY = c(5, 5),
mode = 'canonical')

stim.col <- c("darkblue", "purple", "green4","red3")

# showing only the Y variables, and only those selected in comp 1
cim(res.spls.1level, mapping="Y",
row.sideColors = stim.col[factor(liver.toxicity$treatment[,3])], comp = 1,
#setting up legend:
legend=list(legend = unique(liver.toxicity$treatment[,3]), col=stim.col,
title = "Dose", cex=0.9))


# showing only the X variables, for all selected on comp 1 and 2
cim(res.spls.1level, mapping="X",
row.sideColors = stim.col[factor(liver.toxicity$treatment[,3])],
#setting up legend:
legend=list(legend = unique(liver.toxicity$treatment[,3]), col=stim.col,
title = "Dose", cex=0.9))


# These are the cross correlations between the variables selected in X and Y.
# The similarity matrix is obtained as in our paper in Data Mining
cim(res.spls.1level, mapping="XY")


## End(Not run)

Description

This function generates color-coded Clustered Image Maps (CIMs) ("heat maps") to represent "high-dimensional" data sets analysed with DIABLO.

Usage

cimDiablo(
  object,
  color = NULL,
  color.Y,
  color.blocks,
  comp = NULL,
  margins = c(2, 15),
  legend.position = "topright",
  transpose = FALSE,
  row.names = TRUE,
  col.names = TRUE,
  size.legend = 1.5,
  trim = TRUE,
  ...
)

Arguments

Details

This function is a small wrapper of cim specific to the DIABLO framework.

Value

A list containing the following components:

Author(s)

Amrit Singh, Florian Rohart, Kim-Anh Lê Cao, Al J Abadi

References

Singh A., Shannon C., Gautier B., Rohart F., Vacher M., Tebbutt S. and Lê Cao K.A. (2019), DIABLO: an integrative approach for identifying key molecular drivers from multi-omics assays, Bioinformatics, Volume 35, Issue 17, 1 September 2019, Pages 3055–3062.

Eisen, M. B., Spellman, P. T., Brown, P. O. and Botstein, D. (1998). Cluster analysis and display of genome-wide expression patterns. Proceeding of the National Academy of Sciences of the USA 95, 14863-14868.

Weinstein, J. N., Myers, T. G., O'Connor, P. M., Friend, S. H., Fornace Jr., A. J., Kohn, K. W., Fojo, T., Bates, S. E., Rubinstein, L. V., Anderson, N. L., Buolamwini, J. K., van Osdol, W. W., Monks, A. P., Scudiero, D. A., Sausville, E. A., Zaharevitz, D. W., Bunow, B., Viswanadhan, V. N., Johnson, G. S., Wittes, R. E. and Paull, K. D. (1997). An information-intensive approach to the molecular pharmacology of cancer. Science 275, 343-349.

González I., Lê Cao K.A., Davis M.J., Déjean S. (2012). Visualising associations between paired 'omics' data sets. BioData Mining; 5(1).

mixOmics article:

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

See Also

cim, heatmap, hclust, plotVar, network and

http://mixomics.org/mixDIABLO/ for more details on all options available.

Examples

## default method: shows cross correlation between 2 data sets
#------------------------------------------------------------------
data(nutrimouse)
Y = nutrimouse$diet
data = list(gene = nutrimouse$gene, lipid = nutrimouse$lipid)

nutrimouse.sgccda <- block.splsda(X = data,
                                  Y = Y,
                                  design = "full",
                                  keepX = list(gene = c(10,10), lipid = c(15,15)),
                                  ncomp = 2,
                                  scheme = "centroid")

cimDiablo(nutrimouse.sgccda, comp = c(1,2))
## change trim range
cimDiablo(nutrimouse.sgccda, comp = c(1,2), trim = 4)
## do not trim values
cimDiablo(nutrimouse.sgccda, comp = c(1,2), trim = FALSE)

Description

Displays variable correlation among different blocks

Usage

## S3 method for class 'block.splsda'
circosPlot(
  object,
  comp = 1:min(object$ncomp),
  cutoff,
  color.Y,
  blocks = NULL,
  color.blocks,
  color.cor,
  var.names = NULL,
  showIntraLinks = FALSE,
  line = FALSE,
  size.legend = 0.8,
  ncol.legend = 1,
  size.variables = 0.25,
  size.labels = 1,
  legend = TRUE,
  legend.title = "Expression",
  linkWidth = 1,
  ...
)

## S3 method for class 'block.plsda'
circosPlot(
  object,
  comp = 1:min(object$ncomp),
  cutoff,
  color.Y,
  blocks = NULL,
  color.blocks,
  color.cor,
  var.names = NULL,
  showIntraLinks = FALSE,
  line = FALSE,
  size.legend = 0.8,
  ncol.legend = 1,
  size.variables = 0.25,
  size.labels = 1,
  legend = TRUE,
  legend.title = "Expression",
  linkWidth = 1,
  ...
)

## S3 method for class 'block.spls'
circosPlot(object, ..., group = NULL, Y.name = "Y")

## S3 method for class 'block.pls'
circosPlot(object, ..., group = NULL, Y.name = "Y")

Arguments

Details

circosPlot function depicts correlations of variables selected with block.splsda or block.spls among different blocks, using a generalisation of the method presented in González et al 2012. If ncomp is specified, then only the variables selected on that component are displayed.

The linkWidth argument specifies the width of the links drawn. If a vector of length 2 is provided, the smaller value will correspond to a similarity values designated by cutoff argument, while the larger value will be used for a link with perfect similarity (1), if any.

Value

If saved in an object, the circos plot will output the similarity matrix and the names of the variables displayed on the plot (see attributes(object)).

Author(s)

Michael Vacher, Amrit Singh, Florian Rohart, Kim-Anh Lê Cao, Al J Abadi

References

Singh A., Gautier B., Shannon C., Vacher M., Rohart F., Tebbutt S. and Lê Cao K.A. (2016). DIABLO: multi omics integration for biomarker discovery. BioRxiv available here: http://biorxiv.org/content/early/2016/08/03/067611

mixOmics article:

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

González I., Lê Cao K.A., Davis M.J., Déjean S. (2012). Visualising associations between paired 'omics' data sets. BioData Mining; 5(1).

See Also

block.splsda, references and http://www.mixOmics.org/mixDIABLO for more details.

Examples

data(nutrimouse)
Y = nutrimouse$diet
data = list(gene = nutrimouse$gene, lipid = nutrimouse$lipid)
design = matrix(c(0,1,1,1,0,1,1,1,0), ncol = 3, nrow = 3, byrow = TRUE)


nutrimouse.sgccda <- wrapper.sgccda(X=data,
Y = Y,
design = design,
keepX = list(gene=c(10,10), lipid=c(15,15)),
ncomp = 2,
scheme = "horst")

circosPlot(nutrimouse.sgccda, cutoff = 0.7)
## links widths based on strength of their similarity
circosPlot(nutrimouse.sgccda, cutoff = 0.7, linkWidth = c(1, 10))
## custom legend
circosPlot(nutrimouse.sgccda, cutoff = 0.7, size.legend = 1.1)

## more customisation
circosPlot(nutrimouse.sgccda, cutoff = 0.7, size.legend = 1.1, color.Y = 1:5, 
           color.blocks = c("green","brown"), color.cor = c("magenta", "purple"))
    
par(mfrow=c(2,2))

circosPlot(nutrimouse.sgccda, cutoff = 0.7, size.legend = 1.1)
## also show intra-block correlations
circosPlot(nutrimouse.sgccda, cutoff = 0.7,
           size.legend = 1.1, showIntraLinks = TRUE)
## show lines
circosPlot(nutrimouse.sgccda, cutoff = 0.7, line = TRUE, ncol.legend = 1,
           size.legend = 1.1, showIntraLinks = TRUE)
## custom line legends
circosPlot(nutrimouse.sgccda, cutoff = 0.7, line = TRUE, ncol.legend = 2,
           size.legend = 1.1, showIntraLinks = TRUE)
par(mfrow=c(1,1))

## adjust feature and block names radially
circosPlot(nutrimouse.sgccda, cutoff = 0.7, size.legend = 1.1)
circosPlot(nutrimouse.sgccda, cutoff = 0.7, size.legend = 1.1, 
           var.adj = 0.8, block.labels.adj = -0.5)
## ---  example using breast.TCGA data
data("breast.TCGA")
data = list(mrna = breast.TCGA$data.train$mrna, 
            mirna = breast.TCGA$data.train$mirna, 
            protein = breast.TCGA$data.train$protein)
list.keepX = list(mrna = rep(20, 2), mirna = rep(10,2), protein = c(10, 2))

TCGA.block.splsda = block.splsda(X = data, 
                             Y =breast.TCGA$data.train$subtype,
                             ncomp = 2, keepX = list.keepX, 
                             design = 'full')
circosPlot(TCGA.block.splsda, cutoff = 0.7, line=TRUE)
## show only first 2 blocks
circosPlot(TCGA.block.splsda, cutoff = 0.7, line=TRUE, blocks = c(1,2))
## show only correlations including the mrna block features
circosPlot(TCGA.block.splsda, cutoff = 0.7, blocks.link = 'mrna')

data("breast.TCGA")
data = list(mrna = breast.TCGA$data.train$mrna, mirna = breast.TCGA$data.train$mirna)
list.keepX = list(mrna = rep(20, 2), mirna = rep(10,2))
list.keepY = c(rep(10, 2))

TCGA.block.spls = block.spls(X = data, 
                             Y = breast.TCGA$data.train$protein,
                             ncomp = 2, keepX = list.keepX, 
                             keepY = list.keepY, design = 'full')
circosPlot(TCGA.block.spls, group = breast.TCGA$data.train$subtype, cutoff = 0.7, 
           Y.name = 'protein')
## only show links including mrna
circosPlot(TCGA.block.spls, group = breast.TCGA$data.train$subtype, cutoff = 0.7, 
           Y.name = 'protein', blocks.link = 'mrna')

Description

The functions create a vector of n "contiguous" colors (except the color.mixo which are colors used internally to fit our logo colors).

Usage

color.mixo(num.vector)

color.GreenRed(n, alpha = 1)

color.jet(n, alpha = 1)

color.spectral(n, alpha = 1)

Arguments

Details

The function color.jet(n) create color scheme, beginning with dark blue, ranging through shades of blue, cyan, green, yellow and red, and ending with dark red. This colors palette is suitable for displaying ordered (symmetric) data, with n giving the number of colors desired.

Value

For color.jet(n), color.spectral(n), color.GreenRed(n) a character vector, cv, of color names. This can be used either to create a user-defined color palette for subsequent graphics by palette(cv), a col= specification in graphics functions or in par.

For color.mixo, a vector of colors matching the mixOmics logo (10 colors max.)

Author(s)

Ignacio Gonzalez, Kim-Anh Lê Cao, Benoit Gautier, Al J Abadi

See Also

colorRamp, palette, colors for the vector of built-in "named" colors; hsv, gray, rainbow, terrain.colors, ... to construct colors; and heat.colors, topo.colors for images.

Examples

# -----------------------
# jet colors
# ----------------------
par(mfrow = c(3, 1))
z <- seq(-1, 1, length = 125)
for (n in c(11, 33, 125)) {
image(matrix(z, ncol = 1), col = color.jet(n),
xaxt = 'n', yaxt = 'n', main = paste('n = ', n))
box()
par(usr = c(-1, 1, -1, 1))
axis(1, at = c(-1, 0, 1))
}

## Not run: 
# -----------------------
# spectral colors
# ----------------------
par(mfrow = c(3, 1))
z <- seq(-1, 1, length = 125)
for (n in c(11, 33, 125)) {
image(matrix(z, ncol = 1), col = color.spectral(n),
xaxt = 'n', yaxt = 'n', main = paste('n = ', n))
box()
par(usr = c(-1, 1, -1, 1))
axis(1, at = c(-1, 0, 1))
}

# -----------------------
# GreenRed colors
# ----------------------
par(mfrow = c(3, 1))
z <- seq(-1, 1, length = 125)
for (n in c(11, 33, 125)) {
image(matrix(z, ncol = 1), col = color.GreenRed(n),
xaxt = 'n', yaxt = 'n', main = paste('n = ', n))
box()
par(usr = c(-1, 1, -1, 1))
axis(1, at = c(-1, 0, 1))
}

# # --------------------------------
# mixOmics colors
# # -------------------------------
data(nutrimouse)
X <- nutrimouse$lipid
Y <- nutrimouse$gene
nutri.res <- rcc(X, Y, ncomp = 3, lambda1 = 0.064, lambda2 = 0.008)

my.colors = color.mixo(1:5)
my.pch = ifelse(nutrimouse$genotype == 'wt', 16, 17)
#plotIndiv(nutri.res, ind.names = FALSE, group = my.colors, pch = my.pch, cex = 1.5)

## End(Not run)

Description

The 16S data from the Human Microbiome Project includes only the most diverse bodysites: Antecubital fossa (skin), Stool and Subgingival plaque (oral) and can be analysed using a multilevel approach to account for repeated measurements using our module mixMC. The data include 162 samples (54 unique healthy individuals) measured on 1,674 OTUs.

Usage

data(diverse.16S)

Format

A list containing two data sets, data.TSS and data.raw and some meta data information:

list("data.TSS")

data frame with 162 rows (samples) and 1674 columns (OTUs). The prefiltered normalised data using Total Sum Scaling normalisation.

list("data.raw")

data frame with 162 rows (samples) and 1674 columns (OTUs). The prefiltered raw count OTU data which include a 1 offset (i.e. no 0 values).

list("taxonomy")

data frame with 1674 rows (OTUs) and 6 columns indicating the taxonomy of each OTU.

list("indiv")

data frame with 162 rows indicating sample meta data.

list("bodysite")

factor of length 162 indicating the bodysite with levels "Antecubital_fossa", "Stool" and "Subgingival_plaque".

list("sample")

vector of length 162 indicating the unique individual ID, useful for a multilevel approach to taken into account the repeated measured on each individual.

Details

The data were downloaded from the Human Microbiome Project (HMP, http://hmpdacc.org/HMQCP/all/ for the V1-3 variable region). The original data contained 43,146 OTU counts for 2,911 samples measured from 18 different body sites. We focused on the first visit of each healthy individual and focused on the three most diverse habitats. The prefiltered dataset included 1,674 OTU counts. We strongly recommend to use log ratio transformations on the data.TSS normalised data, as implemented in the PLS and PCA methods, see details on www.mixOmics.org/mixMC.

The data.raw include a 1 offset in order to be log ratios transformed after TSS normalisation. Consequently, the data.TSS are TSS normalisation of data.raw. The CSS normalisation was performed on the orignal data (including zero values)

Value

none

Source

The raw data were downloaded from http://hmpdacc.org/HMQCP/all/. Filtering and normalisation described in our website www.mixOmics.org/mixMC

References

Lê Cao K.-A., Costello ME, Lakis VA, Bartolo, F,Chua XY, Brazeilles R, Rondeau P. MixMC: Multivariate insights into Microbial Communities. PLoS ONE, 11(8): e0160169 (2016).

Description

This function has been renamed tune.rcc, see tune.rcc.

This function has been renamed 'image.tune.rcc', see image.tune.rcc.

This function has been renamed tune.pca.

Value

none

none

none

Description

explained_variance calculates the proportion of variance explained by a set of *orthogonal* variates / components and divides by the total variance in data using the definition of 'redundancy'. This applies to any component-based approaches where components are orthogonal. It is worth noting that any missing values are set to zero (which is the column mean for the centered input data) prior to calculation of total variance in the data. Therefore, this function would underestimate the total variance in presence of abundant missing values. One can use impute.nipals function to impute the missing values to avoid such behaviour.

Usage

explained_variance(data, variates, ncomp)

Arguments

Details

Variance explained by component $t_h$ in $X$ for dimension $h$:

$Rd(X, t_h) = \frac{1}{p} \sum_{j = 1}^p \mbox{cor}^2(X^j, t_h)$

where $X^j$ is the variable centered and scaled, $p$ is the total number of variables.

Value

explained_variance returns a named numeric vector containing the proportion of explained variance for each variate after setting all missing values in the data to zero (see details).

Author(s)

Florian Rohart, Kim-Anh Lê Cao, Al J Abadi

References

Tenenhaus, M., La Régression PLS théorie et pratique (1998). Technip, Paris, chap2.

See Also

spls, splsda, plotIndiv, plotVar, cim, network.

Examples

data(liver.toxicity)
X <- liver.toxicity$gene
Y <- liver.toxicity$clinic

toxicity.spls <- spls(X, Y, ncomp = 2, keepX = c(50, 50), keepY = c(10, 10))

ex = explained_variance(toxicity.spls$X, toxicity.spls$variates$X, ncomp =2)

# ex should be the same as
toxicity.spls$prop_expl_var$X

Description

Create confusion table between a vector of true classes and a vector of predicted classes, calculate the Balanced Error rate

Usage

get.confusion_matrix(truth, all.levels, predicted)

get.BER(confusion)

Arguments

Details

BER is appropriate in case of an unbalanced number of samples per class as it calculates the average proportion of wrongly classified samples in each class, weighted by the number of samples in each class. BER is less biased towards majority classes during the performance assessment.

Value

get.confusion_matrix returns a confusion matrix. get.BER returns the BER from a confusion matrix

Author(s)

Florian Rohart, Al J Abadi

References

mixOmics article:

Rohart F, Gautier B, Singh A, Lê Cao K-A. mixOmics: an R package for 'omics feature selection and multiple data integration. PLoS Comput Biol 13(11): e1005752

See Also

predict.

Examples

data(liver.toxicity)
X <- liver.toxicity$gene
Y <- as.factor(liver.toxicity$treatment[, 4])

## if training is perfomed on 4/5th of the original data
samp <- sample(1:5, nrow(X), replace = TRUE)
test <- which(samp == 1)   # testing on the first fold
train <- setdiff(1:nrow(X), test)

plsda.train <- plsda(X[train, ], Y[train], ncomp = 2)
test.predict <- predict(plsda.train, X[test, ], dist = "max.dist")
Prediction <- test.predict$class$max.dist[, 2]

# the confusion table compares the real subtypes with the predicted subtypes for a 2 component model
confusion.mat = get.confusion_matrix(truth = Y[test],
predicted = Prediction)

get.BER(confusion.mat)

Description

This function provide a image map (checkerboard plot) of the cross-validation score obtained by the tune.rcc function.

Usage

## S3 method for class 'tune.rcc'
image(x, col = heat.colors, ...)

## S3 method for class 'tune.rcc'
plot(x, col = heat.colors, ...)

Arguments

Details

plot.tune.rcc creates an image map of the matrix object$mat containing the cross-validation score obtained by the tune.rcc function. Also a color scales strip is plotted.

Value

none

Author(s)

Sébastien Déjean, Ignacio González, Kim-Anh Le Cao, Al J Abadi

See Also

tune.rcc, image.

Examples

data(nutrimouse)
X <- nutrimouse$lipid
Y <- nutrimouse$gene

## this can take some seconds
cv.score <- tune.rcc(X, Y, validation = "Mfold", plot = FALSE)
plot(cv.score)

# image(cv.score) # same result as plot()

Description

Display two-dimensional visualizations (image maps) of the correlation matrices within and between two data sets.

Usage

imgCor(
  X,
  Y,
  type = "combine",
  X.var.names = TRUE,
  Y.var.names = TRUE,
  sideColors = TRUE,
  interactive.dev = TRUE,
  title = TRUE,
  color,
  row.cex,
  col.cex,
  symkey,
  keysize,
  xlab,
  ylab,
  margins,
  lhei,
  lwid
)

Arguments

Details

If type="combine", the correlation matrix is computed of the combined matrices cbind(X, Y) and then plotted. If type="separate", three correlation matrices are computed, cor(X), cor(Y) and cor(X,Y) and plotted separately on a device. In both cases, a color correlation scales strip is plotted.

The correlation matrices are pre-processed before calling the image function in order to get, as in the numerical representation, the diagonal from upper-left corner to bottom-right one.

Missing values are handled by casewise deletion in the imgCor function.

If X.names = FALSE, the name of each X-variable is hidden. Default value is TRUE.

If Y.names = FALSE, the name of each Y-variable is hidden. Default value is TRUE.

Value

NULL (invisibly)

Author(s)

Ignacio González, Kim-Anh Lê Cao, Florian Rohart, Al J Abadi

See Also

cor, image, color.jet.

Examples

data(nutrimouse)
X <- nutrimouse$lipid
Y <- nutrimouse$gene

## 'combine' type plot (default)
imgCor(X, Y)

## Not run: 
## 'separate' type plot

imgCor(X, Y, type = "separate")

## 'separate' type plot without the name of datas
imgCor(X, Y, X.var.names = FALSE, Y.var.names = FALSE, type = "separate")

## End(Not run)

Description

This function uses nipals function to decompose X into a set of components (t), (pseudo-) singular-values (eig), and feature loadings (p). The original matrix is then approximated/reconstituted using the following equation:

$\hat{X} = t * diag(eig) * t(p)$

The missing values from X are then approximated from this matrix. It is best to ensure enough number of components are used in order to best impute the missing values.

Usage

impute.nipals(X, ncomp, ...)

Arguments

Value

A numeric matrix with missing values imputed.

Author(s)

Al J Abadi

See Also

impute.nipals, pca

Examples

data("nutrimouse")
X <- data.matrix(nutrimouse$lipid)
## add missing values to X to impute and compare to actual values
set.seed(42)
na.ind <- sample(seq_along(X), size = 10)
true.values <- X[na.ind]
X[na.ind] <- NA
X.impute <- impute.nipals(X = X, ncomp = 5)
## compare
round(X.impute[na.ind], 2)
true.values

Description

Performs independent principal component analysis on the given data matrix, a combination of Principal Component Analysis and Independent Component Analysis.

Usage

ipca(
  X,
  ncomp = 2,
  mode = "deflation",
  fun = "logcosh",
  scale = FALSE,
  w.init = NULL,
  max.iter = 200,
  tol = 1e-04
)

Arguments

Details

In PCA, the loading vectors indicate the importance of the variables in the principal components. In large biological data sets, the loading vectors should only assign large weights to important variables (genes, metabolites ...). That means the distribution of any loading vector should be super-Gaussian: most of the weights are very close to zero while only a few have large (absolute) values.

However, due to the existence of noise, the distribution of any loading vector is distorted and tends toward a Gaussian distribtion according to the Central Limit Theroem. By maximizing the non-Gaussianity of the loading vectors using FastICA, we obtain more noiseless loading vectors. We then project the original data matrix on these noiseless loading vectors, to obtain independent principal components, which should be also more noiseless and be able to better cluster the samples according to the biological treatment (note, IPCA is an unsupervised approach).

Algorithm 1. The original data matrix is centered.

2. PCA is used to reduce dimension and generate the loading vectors.

3. ICA (FastICA) is implemented on the loading vectors to generate independent loading vectors.

4. The centered data matrix is projected on the independent loading vectors to obtain the independent principal components.

Value

ipca returns a list with class "ipca" containing the following components:

Author(s)

Fangzhou Yao, Jeff Coquery, Kim-Anh Lê Cao, Florian Rohart, Al J Abadi

References

Yao, F., Coquery, J. and Lê Cao, K.-A. (2011) Principal component analysis with independent loadings: a combination of PCA and ICA. (in preparation)

A. Hyvarinen and E. Oja (2000) Independent Component Analysis: Algorithms and Applications, Neural Networks, 13(4-5):411-430

J L Marchini, C Heaton and B D Ripley (2010). fastICA: FastICA Algorithms to perform ICA and Projection Pursuit. R package version 1.1-13.

See Also

sipca, pca, plotIndiv, plotVar, and http://www.mixOmics.org for more details.

Examples

data(liver.toxicity)

# implement IPCA on a microarray dataset
ipca.res <- ipca(liver.toxicity$gene, ncomp = 3, mode="deflation")
ipca.res

# samples representation
plotIndiv(
    ipca.res,
    ind.names = as.character(liver.toxicity$treatment[, 4]),
    group = as.numeric(as.factor(liver.toxicity$treatment[, 4]))
)

## Not run: 
    plotIndiv(ipca.res,
              cex = 0.01,
              col = as.numeric(as.factor(liver.toxicity$treatment[, 4])),
              style = "3d")

## End(Not run)
# variables representation
plotVar(ipca.res, cex = 0.5)

## Not run: 
plotVar(ipca.res, rad.in = 0.5, cex = 0.5, style="3d")

## End(Not run)

Description

The 16S data come from Koren et al. (2011) and compared the bodysites oral, gut and plaque microbial communities in patients with atherosclerosis. The data can be analysed with our mixMC module. The data include 43 samples measured on 980 OTUs.

Usage

data(Koren.16S)

Format

A list containing two data sets, data.TSS and data.raw and some meta data information:

list("data.TSS")

data frame with 43 rows (samples) and 980 columns (OTUs). The prefiltered normalised data using Total Sum Scaling normalisation.

list("data.raw")

data frame with 43 rows (samples) and 980 columns (OTUs). The prefiltered raw count OTU data which include a 1 offset (i.e. no 0 values).

list("taxonomy")

data frame with 980 rows (OTUs) and 7 columns indicating the taxonomy of each OTU.

list("indiv")

data frame with 43 rows indicating sample meta data.

list("bodysite")

factor of length 43 indicating the bodysite with levels arterial plaque, saliva and stool.

Details

The data are from Koren et al. (2011) who examined the link between oral, gut and plaque microbial communities in patients with atherosclerosis and controls. Only healthy individuals were retained in the analysis. This study contained partially repeated measures from multiple sites including 15 unique patients samples from saliva and stool, and 13 unique patients only sampled from arterial plaque samples and we therefore considered a non multilevel analysis for that experimental design. After prefiltering, the data included 973 OTU for 43 samples. We strongly recommend to use log ratio transformations on the data.TSS normalisd data, as implemented in the PLS and PCA methods, see details on www.mixOmics.org/mixMC.

The data.raw include a 1 offset in order to be log ratios transformed after TSS normalisation. Consequently, the data.TSS are TSS normalisation of data.raw. The CSS normalisation was performed on the orignal data (including zero values)

Value

none

Source

The raw data were downloaded from the QIITA database. Filtering and normalisation described in our website www.mixOmics.org/mixMC

References

Lê Cao K.-A., Costello ME, Lakis VA, Bartolo, F,Chua XY, Brazeilles R, Rondeau P. MixMC: Multivariate insights into Microbial Communities. PLoS ONE, 11(8): e0160169 (2016).

Koren, O., Spor, A., Felin, J., Fak, F., Stombaugh, J., Tremaroli, V., et al.: Human oral, gut, and plaque microbiota in patients with atherosclerosis. Proceedings of the National Academy of Sciences 108(Supplement 1), 4592-4598 (2011)

Description

Three physiological and three exercise variables are measured on twenty middle-aged men in a fitness club.

Usage

data(linnerud)

Format

A list containing the following components:

list("exercise")

data frame with 20 observations on 3 exercise variables.

list("physiological")

data frame with 20 observations on 3 physiological variables.

Value

none

Source

Tenenhaus, M. (1998), Table 1, page 15.

References

Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions Technic.

Description

This data set contains the expression measure of 3116 genes and 10 clinical measurements for 64 subjects (rats) that were exposed to non-toxic, moderately toxic or severely toxic doses of acetaminophen in a controlled experiment.

Usage

data(liver.toxicity)

Format

A list containing the following components:

list("gene")

data frame with 64 rows and 3116 columns. The expression measure of 3116 genes for the 64 subjects (rats).

list("clinic")

data frame with 64 rows and 10 columns, containing 10 clinical variables for the same 64 subjects.

list("treatment")

data frame with 64 rows and 4 columns, containing the treatment information on the 64 subjects, such as doses of acetaminophen and times of necropsies.

list("gene.ID")

data frame with 3116 rows and 2 columns, containing geneBank IDs and gene titles of the annotated genes

Details

The data come from a liver toxicity study (Bushel et al., 2007) in which 64 male rats of the inbred strain Fisher 344 were exposed to non-toxic (50 or 150 mg/kg), moderately toxic (1500 mg/kg) or severely toxic (2000 mg/kg) doses of acetaminophen (paracetamol) in a controlled experiment. Necropsies were performed at 6, 18, 24 and 48 hours after exposure and the mRNA from the liver was extracted. Ten clinical chemistry measurements of variables containing markers for liver injury are available for each subject and the serum enzymes levels are measured numerically. The data were further normalized and pre-processed by Bushel et al. (2007).

Value

none

Source

The two liver toxicity data sets are a companion resource for the paper of Bushel et al. (2007), and was downloaded from:

http://www.biomedcentral.com/1752-0509/1/15/additional/

References

Bushel, P., Wolfinger, R. D. and Gibson, G. (2007). Simultaneous clustering of gene expression data with clinical chemistry and pathological evaluations reveals phenotypic prototypes. BMC Systems Biology 1, Number 15.

Lê Cao, K.-A., Rossouw, D., Robert-Granie, C. and Besse, P. (2008). A sparse PLS for variable selection when integrating Omics data. Statistical Applications in Genetics and Molecular Biology 7, article 35.

Description

This function applies a log transformation to the data, either CLR or ILR

Usage

logratio.transfo(X, logratio = c("none", "CLR", "ILR"), offset = 0)

Arguments

Details

logratio.transfo applies a log transformation to the data, either CLR (centered log ratio transformation) or ILR (Isometric Log Ratio transformation). In the case of CLR log-transformation, X needs to be a matrix of non-negative values and offset is used to shift the values away from 0, as commonly done with counts data.

Value

logratio.transfo simply returns the log-ratio transformed data.

Author(s)

Florian Rohart, Kim-Anh Lê Cao, Al J Abadi

References

Kim-Anh Lê Cao, Mary-Ellen Costello, Vanessa Anne Lakis, Francois Bartolo, Xin-Yi Chua, Remi Brazeilles, Pascale Rondeau mixMC: a multivariate statistical framework to gain insight into Microbial Communities bioRxiv 044206; doi: http://dx.doi.org/10.1101/044206

John Aitchison. The statistical analysis of compositional data. Journal of the Royal Statistical Society. Series B (Methodological), pages 139-177, 1982.

Peter Filzmoser, Karel Hron, and Clemens Reimann. Principal component analysis for compositional data with outliers. Environmetrics, 20(6):621-632, 2009.

See Also

pca, pls, spls, plsda, splsda.

Examples

data(diverse.16S)
CLR = logratio.transfo(X = diverse.16S$data.TSS, logratio = 'CLR')
# no offset needed here as we have put it prior to the TSS, see www.mixOmics.org/mixMC

Description

Converts a matrix in which each row sums to 1 into the nearest matrix of (0,1) indicator variables.

Usage

map(Y)

Arguments

Value

A integer vector with one entry for each row of Y, in which the i-th value is the column index at which the i-th row of Y attains a maximum.

References

C. Fraley and A. E. Raftery (2002). Model-based clustering, discriminant analysis, and density estimation. Journal of the American Statistical Association 97:611-631.

C. Fraley, A. E. Raftery, T. B. Murphy and L. Scrucca (2012). mclust Version 4 for R: Normal Mixture Modeling for Model-Based Clustering, Classification, and Density Estimation. Technical Report No. 597, Department of Statistics, University of Washington.

See Also

unmap

Examples

data(nutrimouse)
Y = unmap(nutrimouse$diet)

map(Y)

Description

This function estimate the rank of a matrix.

Usage

mat.rank(mat, tol)

Arguments

Details

mat.rank estimate the rank of a matrix by computing its singular values $d[i]$ (using nipals). The rank of the matrix can be defined as the number of singular values $d[i] > 0$.

If tol is missing, it is given by tol=max(dim(mat))*max(d)*.Machine$double.eps.

Value

The returned value is a list with components:

Author(s)

Sébastien Déjean, Ignacio González, Al J Abadi

See Also

nipals

Examples

## Hilbert matrix
hilbert <- function(n) { i <- 1:n; 1 / outer(i - 1, i, "+") }
mat <- hilbert(16)
mat.rank(mat)

## Not run: 
## Hilbert matrix with missing data
idx.na <- matrix(sample(c(0, 1, 1, 1, 1), 36, replace = TRUE), ncol = 6)
m.na <- m <- hilbert(9)[, 1:6]
m.na[idx.na == 0] <- NA
mat.rank(m)
mat.rank(m.na)

## End(Not run)

Description

Function to integrate data sets measured on the same samples (N-integration) and to combine multiple independent studies measured on the same variables or predictors (P-integration) using variants of multi-group and generalised PLS (unsupervised analysis).

Usage

mint.block.pls(
  X,
  Y,
  indY,
  study,
  ncomp = 2,
  design,
  scheme,
  mode,
  scale = TRUE,
  init,
  tol = 1e-06,
  max.iter = 100,
  near.zero.var = FALSE,
  all.outputs = TRUE
)

Arguments