Diabetes data

Description

A data.frame with a design and transcriptomic data.

Usage

data(Diabetes)

Details

Clinical study on humans was performed as a 2-way factorial design with two factors both on two levels: bariatric surgery on two levels (before and after the bariatric surgery) and type 2 diabetes (T2D) on two levels (with and without T2D). There were 8 patients without T2D and 7 with T2D. It was discovered that the patients with T2D would be separated in two groups: 3 patients in the group called T2D1 and 4 patients in the group called T2D2. The experiment can therefore also analysed as 2 way factorial design where the disease factor is on three levels. All patients were obese before bariatric surgery (BMI >45). Transcriptome in the subcutaneous adipose tissue were obtained before and one year after bariatric surgery.

Author(s)

Ellen Færgestad Mosleth

References

Dankel et al. 2010. Switch from Stress Response to Homeobox Transcription Factors in Adipose Tissue After Profound Fat Loss. Plos One 5.

Examples

data(Diabetes)
str(Diabetes)

General Effect Modelling

Description

General Effect Modelling

Usage

GEM(formula, data, contrasts = "contr.sum", add_residuals = TRUE, ...)

## S3 method for class 'GEM'
print(x, ...)

## S3 method for class 'GEM'
plot(
  x,
  y = 1,
  what = "raw",
  col = NULL,
  pch = NULL,
  model.line = (what %in% c("raw")),
  ylim = NULL,
  ylab = "",
  xlab = "",
  main = NULL,
  ...
)

tableGEM(object, variable)

## S3 method for class 'GEM'
summary(object, extended = TRUE, df = FALSE, ...)

## S3 method for class 'summary.GEM'
print(x, digits = 2, ...)

Arguments

Value

GEM returns an object of class GEM containing effects, ER values (effect + residuals), fitted values, residuals, features, coefficients, dummy design, symbolic design, dimensions, highest level interaction and feature names.

References

* Mosleth et al. (2021) Cerebrospinal fluid proteome shows disrupted neuronal development in multiple sclerosis. Scientific Report, 11,4087. <doi:10.1038/s41598-021-82388-w>

* E.F. Mosleth et al. (2020). Comprehensive Chemometrics, 2nd edition; Brown, S., Tauler, R., & Walczak, B. (Eds.). Chapter 4.22. Analysis of Megavariate Data in Functional Omics. Elsevier. <doi:10.1016/B978-0-12-409547-2.14882-6>

See Also

Analyses using GEM: elastic, pca, sca, neuralnet, pls. Confidence interval plots: confints. Convenience knock-in and knock-out of effects: knock.in.

Examples

## Multiple Sclerosis
data(MS, package = "gemR")
# Subset to reduce runtime in example
MS$proteins <- MS$proteins[,20:70]

gem <- GEM(proteins ~ group * MS, data = MS)
print(gem)
summary(gem)                                    # Summary of GEM
plot(gem)                                       # Raw data, first feature
plot(gem,2)                                     # Raw data, numbered feature
plot(gem,'Q76L83', col='MS', pch='group')       # Selected colour and plot character
plot(gem,'Q76L83', what='effect MS',
     model.line='effect group')                 # Comparison of factors (points and lines)
print(effs <- colnames(gem$symbolicDesign))     # Inspect factor names
eeffs <- paste0("effect ", effs)
# Example compound plot
old.par <- par(mfrow = c(3,3), mar = c(2,4,4,1))
plot(gem,'Q76L83')                         # Raw data, named feature
plot(gem,'Q76L83', what='fits')            # Fitted values
plot(gem,'Q76L83', what='residuals')       # Residuals
plot(gem,'Q76L83', what=eeffs[1])           # Effect levels
plot(gem,'Q76L83', what=eeffs[2])           # ----||----
plot(gem,'Q76L83', what=eeffs[3])           # ----||----
plot(gem,'Q76L83', what=effs[1])            # ER values
plot(gem,'Q76L83', what=effs[2])            # --------||---------
plot(gem,'Q76L83', what=effs[3])            # --------||---------
par(old.par)

# Complete overview of GEM
tab <- tableGEM(gem, 1)

# In general there can be more than two, effects, more than two levels, and continuous effects:
MS$three <- factor(c(rep(1:3,33),1:2))
gem3    <- GEM(proteins ~ MS * group + three, data = MS)


## Candy assessment
data(candies, package = "HDANOVA")
gemC <- GEM(assessment ~ assessor*candy, data=candies)

# Permutation testing
gemC <- permutation(gemC)
summary(gemC)

# GEM-SCA with ellipsoids in score plots
gemSCA <- sca(gemC)
scoreplot(gemSCA, factor="candy", ellipsoids="confidence")

# GEM-PCA with group colours
gemPCA <- pca(gemC)
scoreplot(gemPCA, factor="candy",
  gr.col=gemPCA$symbolicDesign$candy)


## Lactobacillus
data(Lactobacillus, package = "gemR")
# Subset to reduce runtime in example
Lactobacillus$proteome <- Lactobacillus$proteome[,50:100]

gemLac <- GEM(proteome ~ strain * growthrate, data = Lactobacillus)
print(gemLac)
plot(gemLac)                            # Raw data, first feature
plot(gemLac,2)                          # Raw data, numbered feature
plot(gemLac,'P.LSA0316', col='strain',
    pch='growthrate')                   # Selected colour and plot character
plot(gemLac,'P.LSA0316', what='strain',
    model.line='growthrate')            # Selected model.line


 # Don't run this example, it takes too long
  ## Diabetes
  data(Diabetes, package = "gemR")
  gemDia <- GEM(transcriptome ~ surgery * T2D, data = Diabetes)
  print(gemDia)
  plot(gemDia)                            # Raw data, first feature
  plot(gemDia,2)                          # Raw data, numbered feature
  plot(gemDia,'ILMN_1720829', col='surgery',
      pch='T2D')                          # Selected colour and plot character

Lactobacillus data

Description

A data.frame with a design and proteomic data, transcriptomic data and phenotypic data.

Usage

data(Lactobacillus)

Details

Experiment on Lactobacillus sakei was performed as a 2-way factorial design with two factors both on two levels: strain (L. sakei strains LS25 and 23K) (factor A) and growth condition (high and low glucose availability) (factor B) both on two levels, and their interaction term (factor AB). There were three biological replicates within each group. Transcriptome, proteome and end product profile (lactate, formate, acetate and ethanol) were observed.

Author(s)

Ellen Færgestad Mosleth

References

McLeod et al. 2017. Effects of glucose availability in Lactobacillus sakei; metabolic change and regulation of the proteome and transcriptome. Plos One 12, e0187542.

Examples

data(Lactobacillus)
str(Lactobacillus)

Multiple Sclerosis data

Description

A data.frame with a design and proteomic data.

Usage

data(MS)

Details

Data from biobank are analysed a study population of 101 patients, 37 were diagnosed with multiple sclerosis, and 64 without multiple sclerosis. Of the patients without multiple sclerosis, 50 were diagnosed with other neurological disorders and 14 were neurologically healthy patients who had undergone spinal anaesthesia for orthopaedic surgery on the knee or ankle, i.e. neurologically healthy controls. Unless otherwise stated, all the patients without multiple sclerosis were considered as controls for this study. All patients with multiple sclerosis had relapsing remitting multiple sclerosis. The proteome were obtained on cerebrospinal fluid samples from all patients prior medical treatment for multiple sclerosis. It was discovered the patients separated into two clusters, called group 1 and group 2. This is utilised in the data analysis by considering the data as 2-way factorial design with the two factors: MS and group both on two levels.

Author(s)

Ellen Færgestad Mosleth

References

* Opsahl, J.A. et al. Label-free analysis of human cerebrospinal fluid addressing various normalization strategies and revealing protein groups affected by multiple sclerosis. Proteomics 16, 1154-1165 (2016).

* Ellen Færgestad Mosleth, Christian Alexander Vedeler, Kristian Hovde Liland, Anette McLeod, Gerd Haga Bringland, Liesbeth Kroondijk, Frode Berven, Artem Lysenko, Christopher J. Rawlings, Karim El-Hajj Eid, Jill Anette Opsahl, Bjørn Tore Gjertsen, Kjell-Morten Myhr and Sonia Gavasso, Cerebrospinal fluid proteome shows disrupted neuronal development in multiple sclerosis. Scientific Reports – Nature 11(4087), (2021).

Examples

data(MS)
str(MS)

Confidence Intervals of Effect Differences

Description

Confidence Intervals of Effect Differences

Usage

confints(X1, ...)

## Default S3 method:
confints(X1, X2, confidence = 0.95, df.used = 0, ...)

## S3 method for class 'GEM'
confints(
  X1,
  factor = 1,
  levels = c(1, 2),
  confidence = 0.95,
  df.used = X1$df.used,
  ...
)

## S3 method for class 'confints'
plot(
  x,
  y,
  xlab = "",
  ylab = "values",
  sorted = TRUE,
  labels = FALSE,
  nonZero = FALSE,
  xlim = NULL,
  ylim = NULL,
  text.pt = 12,
  ...
)

Arguments

Value

An object of class confints, which holds the information needed to perform statistics or plot the confidence intervals is returned from confints. The plotting routine returns a ggplot structure for plotting.

See Also

Analyses using GEM: elastic, pca, sca, neuralnet, pls.

Examples

data(MS)
# Subset to reduce runtime in example
MS$proteins <- MS$proteins[,20:70]

# Compare MS and non-MS patients within group 1
conf <- with(MS, confints(proteins[MS == "yes" & group == 1,],
                          proteins[MS == "no"  & group == 1,]))
p1 <- plot(conf)
p2 <- plot(conf, nonZero = TRUE) # Only intervals without 0.
grid.arrange(p1,p2)

# Comparison repeated but based on ER matrices
gem <- GEM(proteins ~ MS * group, data = MS)
print(effs <- colnames(gem$symbolicDesign)) # Inspect factor names
confGEM <- confints(gem, factor=effs[3], levels=c("yes.1","no.1"))
p1g <- plot(confGEM)
p2g <- plot(confGEM, nonZero = TRUE) # Only intervals without 0.
grid.arrange(p1g,p2g)

# Shorter plot with labels
confShort <- conf[1:10,]
p1 <- plot(confShort, labels = TRUE)
p2 <- plot(confShort, labels = TRUE, nonZero = TRUE)
grid.arrange(p1,p2)

Elastic-net modeling of GEM objects.

Description

Elastic-net modeling of GEM objects.

Usage

elastic(gem, ...)

## S3 method for class 'GEM'
elastic(
  gem,
  effect,
  alpha = 0.5,
  newdata = NULL,
  validation,
  segments = NULL,
  measure = measure,
  family = family,
  ...
)

Arguments

Value

An object of class GEMglmnet, cv.glmnet, list containing the fitted Elastic-net model, classifications/predictions and data.

See Also

Analyses using GEM: pca, sca, neuralnet, pls. Confidence interval plots: confints. Convenience knock-in and knock-out of effects: knock.in.

Examples

## Multiple Sclerosis data
data(MS, package = "gemR")
# Subset to reduce runtime in example
MS$proteins <- MS$proteins[,20:70]

gem <- GEM(proteins ~ MS * group, data = MS)
elasticMod <- elastic(gem, 'MS', validation = "CV")
sum(elasticMod$classes == MS$MS)
plot(elasticMod)            # Model fit
plot(elasticMod$glmnet.fit) # Coefficient trajectories

# Select all proteins with non-zeros coefficients
coefs     <- coef(elasticMod)
(selected <- names(which(coefs[,1] != 0)))

 # Time consuming due to many variables
  ## Diabetes data
  data(Diabetes, package = "gemR")
  gem.Dia <- GEM(transcriptome ~ surgery * T2D, data = Diabetes)
  elasticMod <- elastic(gem.Dia, 'T2D', validation = "LOO")

Knock-in and knock-out of effect(s)

Description

Convenience functions to apply to GEM objects to isolate/extract (knock-in) one or more effects (and possibly residuals) or to remove (knock-out) one or more effects.

Usage

knock.in(object, effect, residuals = TRUE)

knock.out(object, effect, residuals = TRUE)

Arguments

Value

A data.frame of ER values where effects have been knocked in (isolated/extracted from data) or knocked out (removed from data).

Examples

data(MS, package = "gemR")
gem <- GEM(proteins ~ MS * group, data = MS)

# Extract interaction between 'MS' and 'group
ER.isolated <- knock.in(gem, 'MS:group')

# Remove main effect of 'group'
ER.cleaned <- knock.out(gem, 'group')

Neural Network by Multilayer Perceptron

Description

Neural Network by Multilayer Perceptron

Usage

neuralnet(
  object,
  formula,
  factor = 1,
  hidden = c(2),
  linear.output = FALSE,
  ...
)

Arguments

Value

A neuralnet object that can be inspected and plotted.

See Also

Analyses using GEM: elastic, pca, sca, neuralnet, pls. Confidence interval plots: confints. Convenience knock-in and knock-out of effects: knock.in.

Examples

data(candies, package = "HDANOVA")
gemC <- GEM(assessment ~ assessor*candy, data=candies)

# Neural network model
nn <- neuralnet(gemC, factor = "candy", hidden = c(2))
plot(nn, rep="best")

# Network weights (input and hidden layers)
nn$weights

Principal Component Analysis

Description

This function performs Principal Component Analysis (SCA) on a GEM/hdanova object.

Usage

pca(object)

Arguments

Value

An updated GEM/hdanova object with PCA results.

See Also

Analyses using GEM: elastic, pca, sca, neuralnet, pls. Confidence interval plots: confints. Convenience knock-in and knock-out of effects: knock.in.

Examples

# Load candies data
data(candies, package="HDANOVA")

# Basic HDANOVA model with two factors
mod <- GEM(assessment ~ candy + assessor, data=candies)
mod <- pca(mod)
scoreplot(mod)

Plot function and extraction method for GEM-based PLS

Description

Plot function and extraction method for GEM-based PLS

Usage

## S3 method for class 'GEMpls'
plot(x, y, ylab = "error", xlab = "nvar", main = "Shaving", ...)

## S3 method for class 'GEMpls'
print(x, ...)

## S3 method for class 'GEMpls'
summary(object, ...)

scores(object, ...)

scoreplot(object, ...)

loadings(object, ...)

loadingplot(object, ...)

corrplot(object, ...)

R2(object, ...)

mvrValstats(object, ...)

explvar(object, ...)

Arguments

Value

A plot of the PLS model's cross-validated accuracy or the Shaving results if available.

See Also

pls has examples of how to use this function.

Partial Least Squares modelling of GEM objects.

Description

The output of GEM is used as input to a PLS classification with the selected effect as response. It is possible to compare two models using the gem2 argument. Variable selection is available through Jackknifing (from package pls) and Shaving (from package plsVarSel).

Usage

pls(gem, ...)

## S3 method for class 'GEM'
pls(
  gem,
  effect,
  ncomp,
  newdata = NULL,
  gem2,
  validation,
  jackknife = NULL,
  shave = NULL,
  df.used = gem$df.used,
  ...
)

Arguments

Details

If using the shave options, the segment type is given as type instead of segment.type (see examples).

Value

An object of class GEMpls, mvr, list containing the fitted PLS model, classifications/predictions, data and optionally Jackknife or Shaving results.

See Also

Analyses using GEM: elastic, pca, sca, neuralnet. Confidence interval plots: confints. Convenience knock-in and knock-out of effects: knock.in.

Examples

data(MS, package = "gemR")
# Subset to reduce runtime in example
MS$proteins <- MS$proteins[,20:70]

gem <- GEM(proteins ~ MS * group, data = MS[-1,])

# Simple PLS using interleaved cross-validation
plsMod <- pls(gem, 'MS', 6, validation = "CV",
              segment.type = "interleaved", length.seg = 25)
plot(plsMod)
scoreplot(plsMod, labels = "names")

# PLS with shaving of variables (mind different variable for cross-validation type)
plsModS <- pls(gem, 'MS', 6, validation = "CV",
              type = "interleaved", length.seg=25, shave = TRUE)
# Error as a function of remaining variables
plot(plsModS)
# Selected variables for minimum error
with(plsModS$shave, colnames(X)[variables[[min.red+1]]])

 # Time consuming due to leave-one-out cross-validation
  plsModJ <- pls(gem, 'MS', 5, validation = "LOO",
              jackknife = TRUE)
  colSums(plsModJ$classes == as.numeric(MS$MS[-1]))
  # Jackknifed coefficient P-values (sorted)
  plot(sort(plsModJ$jack[,1,1]), pch = '.', ylab = 'P-value')
  abline(h=c(0.01,0.05),col=2:3)

  scoreplot(plsModJ)
  scoreplot(plsModJ, comps=c(1,3))   # Selected components
  # Use MS categories for colouring and clusters for plot characters.
  scoreplot(plsModJ, col = gem$symbolicDesign[['MS']],
                  pch = 20+as.numeric(gem$symbolicDesign[['group']]))
  loadingplot(plsModJ, scatter=TRUE) # scatter=TRUE for scatter plot

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

HDANOVA

permutation

gridExtra

grid.arrange

Simultaenous Component Analysis

Description

This function performs Simultaneous Component Analysis (SCA) on a hdanova object.

Usage

sca(object)

Arguments

Value

An updated hdanova object with SCA results.

See Also

Analyses using GEM: elastic, pca, sca, neuralnet, pls. Confidence interval plots: confints. Convenience knock-in and knock-out of effects: knock.in.

Examples

# Load candies data
data(candies, package="HDANOVA")

# Basic HDANOVA model with two factors
mod <- GEM(assessment ~ candy + assessor, data=candies)
mod <- sca(mod)
scoreplot(mod)