Package 'pRoloc'

Description

Adds GO annotations to the feature data

Usage

addGoAnnotations(
  object,
  params,
  evidence,
  useID = FALSE,
  fcol = "GOAnnotations",
  ...
)

Arguments

Value

An updated MSnSet with new feature data column called GOAnnotations containing a matrix of GO annotations

Author(s)

Lisa M Breckels

Examples

library(pRolocdata)
data(dunkley2006)
par <- setAnnotationParams(inputs =
                   c("Arabidopsis thaliana genes",
                   "Gene stable ID"))
## add protein sets/annotation information
xx <- addGoAnnotations(dunkley2006, par)
dim(fData(xx)$GOAnnotations)

## filter sets
xx <- filterMinMarkers(xx, n = 50)
dim(fData(xx)$GOAnnotations)
xx <- filterMaxMarkers(xx, p = .25)
dim(fData(xx)$GOAnnotations)

## Subset for specific protein sets
sub <- subsetMarkers(xx, keep = c("vacuole"))

## Order protein sets
res <- orderGoAnnotations(xx, k = 1:3, p = 1/3, verbose = FALSE)
if (interactive()) {
pRolocVis(res, fcol = "GOAnnotations")
}

Description

Adds a legend to a plot2D figure.

Usage

addLegend(
  object,
  fcol = "markers",
  where = c("bottomleft", "bottom", "bottomright", "left", "topleft", "top", "topright",
    "right", "center", "other"),
  col,
  bty = "n",
  ...
)

Arguments

Details

The function has been updated in version 1.3.6 to recycle the default colours when more organelle classes are provided. See plot2D for details.

Value

Invisibly returns NULL

Author(s)

Laurent Gatto

Description

The function adds a 'markers' feature variable. These markers are read from a comma separated values (csv) spreadsheet file. This markers file is expected to have 2 columns (others are ignored) where the first is the name of the marker features and the second the group label. Alternatively, a markers named vector as provided by the pRolocmarkers function can also be used.

Usage

addMarkers(object, markers, mcol = "markers", fcol, verbose = TRUE)

Arguments

Details

It is essential to assure that featureNames(object) (or fcol, see below) and marker names (first column) match, i.e. the same feature identifiers and case fold are used.

Value

A new instance of class MSnSet with an additional markers feature variable.

Author(s)

Laurent Gatto

See Also

See pRolocmarkers for a list of spatial markers and markers for details about markers encoding.

Examples

library("pRolocdata")
data(dunkley2006)
atha <- pRolocmarkers("atha")
try(addMarkers(dunkley2006, atha)) ## markers already exists
fData(dunkley2006)$markers.org <- fData(dunkley2006)$markers
fData(dunkley2006)$markers <- NULL
marked <- addMarkers(dunkley2006, atha)
fvarLabels(marked)
## if 'makers' already exists
marked <- addMarkers(marked, atha, mcol = "markers2")
fvarLabels(marked)
stopifnot(all.equal(fData(marked)$markers, fData(marked)$markers2))
plot2D(marked)
addLegend(marked, where = "topleft", cex = .7)

Description

Class to store annotation parameters to automatically query a Biomart server, retrieve relevant annotation for a set of features of interest using, for example getGOFromFeatures and makeGoSet.

Objects from the Class

Objects can be created and set with the setAnnotationParams function. Object are created by calling without any arguments setAnnotationParams(), which will open an interactive interface. Depending on the value of "many.graphics" option, a graphical of a text-based menu will open (the text interface can be forced by setting the graphics argument to FALSE: setAnnotationParams(graphics = FALSE)). The menu will allow to select the species of interest first and the type of features (ENSEMBL gene identifier, Entrez id, ...) second.

The species that are available are those for which ENSEMBL data is available in Biomart and have a set of attributes of interest available. The compatible identifiers for downstream queries are then automatically filtered and displayed for user selection.

It is also possible to pass a parameter inputs, a character vector of length 2 containing a pattern uniquely matching the species of interest (in position 1) and a patterns uniquely matching the feature types (in position 2). If the matches are not unique, an error will be thrown.

A new instance of the AnnotationParams will be created to enable easy and automatic query of the Mart instance. The instance is invisibly returned and stored in a global variable in the pRoloc package's private environment for automatic retrieval. If a variable containing an AnnotationParams instance is already available, it can be set globally by passing it as argument to the setAnnotationParams function. Globally set AnnotationParams instances can be accessed with the getAnnotationParams function.

See the pRoloc-theta vignette for details.

Slots

mart:

Object of class "Mart" from the biomaRt package.

martname:

Object of class "character" with the name of the mart instance.

dataset:

Object of class "character" with the data set of the mart instance.

filter:

Object of class "character" with the filter to be used when querying the mart instance.

date:

Object of class "character" indicating when the current instance was created.

biomaRtVersion:

Object of class "character" with the biomaRt version used to create the AnnotationParams instance.

.__classVersion__:

Object of class "Versions" with the version of the AnnotationParams class of the current instance.

Methods

show

signature(object = "AnnotationParams"): to display objects.

Author(s)

Laurent Gatto <[email protected]>

See Also

getGOFromFeatures, makeGoSet and the pRoloc-theta vignette.

Examples

data(andy2011params)
andy2011params
data(dunkley2006params)
dunkley2006params

try(setAnnotationParams(inputs = c("nomatch1", "nomatch2")))
setAnnotationParams(inputs = c("Human genes",
			       "UniProtKB/Swiss-Prot ID"))
getAnnotationParams()

Description

Checks the marker and unknown feature overlap of two MSnSet instances.

Usage

checkFeatureNamesOverlap(x, y, fcolx = "markers", fcoly, verbose = TRUE)

Arguments

Value

Invisibly returns a named list of common markers, unique x markers, unique y markers in, common unknowns, unique x unknowns and unique y unknowns.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(andy2011)
data(andy2011goCC)
checkFeatureNamesOverlap(andy2011, andy2011goCC)
featureNames(andy2011goCC)[1] <- "ABC"
res <- checkFeatureNamesOverlap(andy2011, andy2011goCC)
res$markersX
res$markersY

Description

Extracts qualitative feature variables from two MSnSet instances and compares with a contingency table.

Usage

checkFvarOverlap(x, y, fcolx = "markers", fcoly, verbose = TRUE)

Arguments

Value

Invisibly returns a named list with the values of the diagonal, upper and lower triangles of the contingency table.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(dunkley2006)
res <- checkFvarOverlap(dunkley2006, dunkley2006,
                        "markers", "markers.orig")
str(res)

Description

In the original protein correlation profiling (PCP), Andersen et al. use the peptide normalised profiles along gradient fractions and compared them with the reference profiles (or set of profiles) by computing $Chi^2$ values, $\frac{\sum (x_i - x_p)^2}{x_p}$, where $x_i$ is the normalised value of the peptide in fraction i and $x_p$ is the value of the marker (from Wiese et al., 2007). The protein $Chi^2$ is then computed as the median of the peptide $Chi^2$ values. Peptides and proteins with similar profiles to the markers will have small $Chi^2$ values.

The chi2 methods implement this idea and compute such Chi^2 values for sets of proteins.

Methods

signature(x = "matrix", y = "matrix", method = "character", fun = "NULL", na.rm = "logical")

Compute nrow(x) times nrow(y) $Chi^2$ values, for each x, y feature pair. Method is one of "Andersen2003" or "Wiese2007"; the former (default) computed the $Chi^2$ as sum(y-x)^2/length(x), while the latter uses sum((y-x)^2/x). na.rm defines if missing values (NA and NaN) should be removed prior to summation. fun defines how to summarise the $Chi^2$ values; default, NULL, does not combine the $Chi^2$ values.

signature(x = "matrix", y = "numeric", method = "character", na.rm = "logical")

Computes nrow(x) $Chi^2$ values, for all the $(x_i, y)$ pairs. See above for the other arguments.

signature(x = "numeric", y = "matrix", method = "character", na.rm = "logical")

Computes nrow(y) $Chi^2$ values, for all the $(x, y_i)$ pairs. See above for the other arguments.

signature(x = "numeric", y = "numeric", method = "character", na.rm = "logical")

Computes the $Chi^2$ value for the $(x, y)$ pairs. See above for the other arguments.

Author(s)

Laurent Gatto <[email protected]>

References

Andersen, J. S., Wilkinson, C. J., Mayor, T., Mortensen, P. et al., Proteomic characterization of the human centrosome by protein correlation profiling. Nature 2003, 426, 570 - 574.

Wiese, S., Gronemeyer, T., Ofman, R., Kunze, M. et al., Proteomics characterization of mouse kidney peroxisomes by tandem mass spectrometry and protein correlation profiling. Mol. Cell. Proteomics 2007, 6, 2045 - 2057.

See Also

empPvalues

Examples

mrk <- rnorm(6)
prot <- matrix(rnorm(60), ncol = 6)
chi2(mrk, prot, method = "Andersen2003")
chi2(mrk, prot, method = "Wiese2007")

pepmark <- matrix(rnorm(18), ncol = 6)
pepprot <- matrix(rnorm(60), ncol = 6)
chi2(pepmark, pepprot)
chi2(pepmark, pepprot, fun = sum)

Description

Calculates class weights to be used for parameter optimisation and classification such as svmOptimisation or svmClassification - see the pRoloc tutorial vignette for an example. The weights are calculated for all non-unknown classes the inverse of the number of observations.

Usage

classWeights(object, fcol = "markers")

Arguments

Value

A table of class weights

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(hyperLOPIT2015)
classWeights(hyperLOPIT2015)
data(dunkley2006)
classWeights(dunkley2006)

Description

This function computes the mean (normalised) pairwise distances for pre-defined sets of proteins.

Usage

clustDist(object, k = 1:5, fcol = "GOAnnotations", n = 5, verbose = TRUE, seed)

Arguments

Details

The input to the function is a MSnSet dataset containing a matrix appended to the feature data slot identifying the membership of protein instances to a pre-defined set(s) e.g. a specific Gene Ontology term etc.

For each protein set, the clustDist function (i) extracts all instances belonging to the set, (ii) using the kmeans algorithm fits and tests k = c(1:5) (default) cluster components to each set, (iii) calculates the mean pairwise distance for each k tested.

Note: currently distances are calcualted in Euclidean space, but other distance metrics will be supported in the future).

The output is a list of ClustDist objects, one per information cluster. The ClustDist class summarises the algorithm information such as the number of k's tested for the kmeans, and mean and normalised pairwise Euclidean distances per numer of component clusters tested. See ?ClustDist for more details.

Value

An instance of "ClustDistList" containing a "ClustDist" instance for every protein set, which summarises the algorithm information such as the number of k's tested for the kmeans, and mean and normalised pairwise Euclidean distances per numer of component clusters tested.

Author(s)

Lisa Breckels

See Also

For class definitions see "ClustDistList" and "ClustDist".

Examples

library(pRolocdata)
data(dunkley2006)
par <- setAnnotationParams(inputs =
                   c("Arabidopsis thaliana genes",
                   "Gene stable ID"))
## add protein sets/annotation information
xx <- addGoAnnotations(dunkley2006, par)
## filter
xx <- filterMinMarkers(xx, n = 50)
xx <- filterMaxMarkers(xx, p = .25)
## get distances for protein sets 
dd <- clustDist(xx)
## plot clusters for first 'ClustDist' object 
## in the 'ClustDistList'
plot(dd[[1]], xx)
## plot distances for all protein sets 
plot(dd)
## Extract normalised distances
## Normalise by n^1/3
minDist <- getNormDist(dd, p = 1/3)
## Get new order according to lowest distance
o <- order(minDist)
## Re-order GOAnnotations 
fData(xx)$GOAnnotations <- fData(xx)$GOAnnotations[, o]
if (interactive()) {
pRolocVis(xx, fcol = "GOAnnotations")
}

Description

The ClustDist summaries algorithm information, from running the clustDist function, such as the number of k's tested for the kmeans, and mean and normalised pairwise (Euclidean) distances per numer of component clusters tested.

Objects from the Class

Object of this class are created with the clustDist function.

Slots

k:

Object of class "numeric" storing the number of k clusters tested.

dist:

Object of class "list" storing the list of distance matrices.

term:

Object of class "character" describing GO term name.

id:

Object of class "character" describing the GO term ID.

nrow:

Object of class "numeric" showing the number of instances in the set

clustsz:

Object of class "list" describing the number of instances for each cluster for each k tested

components:

Object of class "vector" storing the class membership of each protein for each k tested.

fcol:

Object of class "character" showing the feature column name in the corresponding MSnSet where the protein set information is stored.

Methods

plot

Plots the kmeans clustering results.

show

Shows the object.

Author(s)

Lisa M Breckels <[email protected]>

Examples

showClass("ClustDist")
  
  library('pRolocdata')
  data(dunkley2006)
  par <- setAnnotationParams(inputs =
                    c("Arabidopsis thaliana genes",
                      "Gene stable ID"))
                    
  ## add protein set/annotation information                  
  xx <- addGoAnnotations(dunkley2006, par)
  
  ## filter
  xx <- filterMinMarkers(xx, n = 50)
  xx <- filterMaxMarkers(xx, p = .25)
  
  ## get distances for protein sets
  dd <- clustDist(xx)
  
  ## plot clusters for first 'ClustDist' object 
  ## in the 'ClustDistList'
  plot(dd[[1]], xx)
  
  ## plot distances for all protein sets 
  plot(dd)

Description

A class for storing lists of ClustDist instances.

Objects from the Class

Object of this class are created with the clustDist function.

Slots

x:

Object of class list containing valid ClustDist instances.

log:

Object of class list containing an object creation log, containing among other elements the call that generated the object.

.__classVersion__:

The version of the instance. For development purposes only.

Methods

"[["

Extracts a single ClustDist at position.

"["

Extracts one of more ClustDists as ClustDistList.

length

Returns the number of ClustDists.

names

Returns the names of ClustDists, if available. The replacement method is also available.

show

Display the object by printing a short summary.

lapply(x, FUN, ...)

Apply function FUN to each element of the input x. If the application of FUN returns and ClustDist, then the return value is an ClustDistList, otherwise a list

.

plot

Plots a boxplot of the distance results per protein set.

Author(s)

Lisa M Breckels <[email protected]>

Examples

library('pRolocdata')
  data(dunkley2006)
  par <- setAnnotationParams(inputs =
                    c("Arabidopsis thaliana genes",
                    "Gene stable ID"))

  ## add protein set/annotation information
  xx <- addGoAnnotations(dunkley2006, par)

  ## filter
  xx <- filterMinMarkers(xx, n = 50)
  xx <- filterMaxMarkers(xx, p = .25)

  ## get distances for protein sets
  dd <- clustDist(xx)

  ## plot distances for all protein sets
  plot(dd)

  names(dd)

  ## Extract a sub-list of ClustDist objects
  dd[1]

  ## Extract 1st ClustDist object
  dd[[1]]

Description

Andersen et al. (2003) used a fixed $Chi^2$ threshold of 0.05 to identify organelle-specific candidates. This function computes empirical p-values by permutation the markers relative intensities and computed null $Chi^2$ values.

Usage

empPvalues(marker, corMatrix, n = 100, ...)

Arguments

Value

A numeric of length nrow(corMatrix).

Author(s)

Laurent Gatto <[email protected]>

References

Andersen, J. S., Wilkinson, C. J., Mayor, T., Mortensen, P. et al., Proteomic characterization of the human centrosome by protein correlation profiling. Nature 2003, 426, 570 - 574.

See Also

chi2 for $Chi^2$ calculation.

Examples

set.seed(1)
mrk <- rnorm(6, 5, 1)
prot <- rbind(matrix(rnorm(120, 5, 1), ncol = 6),
              mrk + rnorm(6))
mrk <- mrk/sum(mrk)
prot <- prot/rowSums(prot)
empPvalues(mrk, prot)

Description

This function replaces a string or regular expression in a feature variable using the sub function.

Usage

fDataToUnknown(object, fcol = "markers", from = "^$", to = "unknown", ...)

Arguments

Value

An updated MSnSet.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(dunkley2006)
getMarkers(dunkley2006, "markers")
dunkley2006 <- fDataToUnknown(dunkley2006,
                              from = "unknown", to = "unassigned")
getMarkers(dunkley2006, "markers")

Description

Removes columns or rows that have a certain proportion or absolute number of 0 values.

Usage

filterBinMSnSet(object, MARGIN = 2, t, q, verbose = TRUE)

Arguments

Value

A filtered MSnSet.

Author(s)

Laurent Gatto

See Also

zerosInBinMSnSet, filterZeroCols, filterZeroRows.

Examples

set.seed(1)
m <- matrix(sample(0:1, 25, replace=TRUE), 5)
m[1, ] <- 0
m[, 1] <- 0
rownames(m) <- colnames(m) <- letters[1:5]
fd <- data.frame(row.names = letters[1:5])
x <- MSnSet(exprs = m, fData = fd, pData = fd)
exprs(x)
## Remove columns with no 1s
exprs(filterBinMSnSet(x, MARGIN = 2, t = 0))
## Remove columns with one 1 or less
exprs(filterBinMSnSet(x, MARGIN = 2, t = 1))
## Remove columns with two 1s or less
exprs(filterBinMSnSet(x, MARGIN = 2, t = 2))
## Remove columns with three 1s 
exprs(filterBinMSnSet(x, MARGIN = 2, t = 3))
## Remove columns that have half or less of 1s
exprs(filterBinMSnSet(x, MARGIN = 2, q = 0.5))

Description

Removes annotation information that contain more that a certain number/percentage of proteins

Usage

filterMaxMarkers(object, n, p = 0.2, fcol = "GOAnnotations", verbose = TRUE)

Arguments

Value

An updated MSnSet

See Also

addGoAnnotations and example therein.

Description

Removes annotation information that contain less that a certain number/percentage of proteins

Usage

filterMinMarkers(object, n = 10, p, fcol = "GOAnnotations", verbose = TRUE)

Arguments

Value

An updated MSnSet.

Author(s)

Lisa M Breckels

See Also

addGoAnnotations and example therein.

Description

Removes all assay data columns/rows that are composed of only 0, i.e. have a colSum/rowSum of 0.

Usage

filterZeroCols(object, verbose = TRUE)

filterZeroRows(object, verbose = TRUE)

Arguments

Value

An MSnSet.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(andy2011goCC)
any(colSums(exprs(andy2011goCC)) == 0)
exprs(andy2011goCC)[, 1:5] <- 0
ncol(andy2011goCC)
ncol(filterZeroCols(andy2011goCC))

Description

Regularisation framework containers.

Objects from the Class

Object of this class are created with the respective regularisation function: knnOptimisation, svmOptimisation, plsdaOptimisation, knntlOptimisation, ...

Slots

algorithm:

Object of class "character" storing the machine learning algorithm name.

hyperparameters:

Object of class "list" with the respective algorithm hyper-parameters tested.

design:

Object of class "numeric" describing the cross-validation design, the test data size and the number of replications.

log:

Object of class "list" with warnings thrown during the hyper-parameters regularisation.

seed:

Object of class "integer" with the random number generation seed.

results:

Object of class "matrix" of dimenstions times (see design) by number of hyperparameters + 1 storing the macro F1 values for the respective best hyper-parameters for each replication.

f1Matrices:

Object of class "list" with respective times cross-validation F1 matrices.

cmMatrices:

Object of class "list" with respective times contingency matrices.

testPartitions:

Object of class "list" with respective times test partitions.

datasize:

Object of class "list" with details about the respective inner and outter training and testing data sizes.

Only in ThetaRegRes:

predictions:

A list of predictions for the optimisation iterations.

otherWeights:

Alternative best theta weigts: a vector per iterations, NULL if no other best weights were found.

Methods

getF1Scores

Returns a matrix of F1 scores for the optimisation parameters.

f1Count

signature(object = "GenRegRes", t = "numeric") and signature(object = "ThetaRegRes", t = "numeric"): Constructs a table of all possible parameter combination and count how many have an F1 scores greater or equal than t. When t is missing (default), the best F1 score is used. This method is useful in conjunctin with plot.

getParams

Returns the best parameters. It is however strongly recommended to inspect the optimisation results. For a ThetaRegRes optimisation result, the method to chose the best parameters can be "median" (default) or "mean" (the median or mean of the best weights is chosen), "max" (the first weights with the highest macro-F1 score, considering that multiple max scoring combinations are possible) or "count" (the observed weight that get the maximum number of observations, see f1Count). The favourP argument can be used to prioritise weights that favour the primary data (i.e. heigh weights). See favourPrimary below.

getSeed

Returns the seed used for the optimisation run.

getWarnings

signature(object = "GenRegRes"): Returns a vector of recorded warnings.

levelPlot

signature(object = "GenRegRes"): Plots a heatmap of of the optimisation results. Only for "GenRegRes" instances.

plot

Plots the optisisation results.

show

Shows the object.

Other functions

Only for ThetaRegRes:

combineThetaRegRes(object)

Takes a list of ThetaRegRes instances to be combined and returnes a new ThetaRegRes instance.

favourPrimary(primary, auxiliary, object, verbose = TRUE)

Takes the primary and auxiliary data sources (two MSnSet instances) and a ThetaRegRes object and returns and updated ThetaRegRes instance containing best parameters/weigths (see the getParams function) favouring the primary data when multiple best theta weights are available.

Author(s)

Laurent Gatto <[email protected]>

Examples

showClass("GenRegRes")
showClass("ThetaRegRes")

Description

The function pulls the gene ontology (GO) terms for a set of feature names.

Usage

getGOFromFeatures(
  id,
  namespace = "cellular_component",
  evidence = NULL,
  params = NULL,
  verbose = FALSE,
  nmax = 500
)

Arguments

Value

A data.frame with relevant GO terms.

Author(s)

Laurent Gatto

Examples

library(pRolocdata)
data(dunkley2006)
data(dunkley2006params)
dunkley2006params
fn <- featureNames(dunkley2006)[1:5]
getGOFromFeatures(fn, params = dunkley2006params)

Description

Convenience accessor to the organelle classes in an 'MSnSet'. This function returns the organelle classes of an MSnSet instance. As a side effect, it prints out the classes.

Usage

getMarkerClasses(object, fcol = "markers", ...)

Arguments

Value

A character vector of the organelle classes in the data.

Author(s)

Lisa Breckels and Laurent Gatto

See Also

getMarkers to extract the marker proteins. See markers for details about spatial markers storage and encoding.

Examples

library("pRolocdata")
data(dunkley2006)
organelles <- getMarkerClasses(dunkley2006)
## same if markers encoded as a matrix
dunkley2006 <- mrkVecToMat(dunkley2006, mfcol = "Markers")
organelles2 <- getMarkerClasses(dunkley2006, fcol = "Markers")
stopifnot(all.equal(organelles, organelles2))

Description

Convenience accessor to the organelle markers in an MSnSet. This function returns the organelle markers of an MSnSet instance. As a side effect, it print out a marker table.

Usage

getMarkers(object, fcol = "markers", names = TRUE, verbose = TRUE)

Arguments

Value

A character (matrix) of length (ncol) ncol(object), depending on the vector or matrix encoding of the markers.

Author(s)

Laurent Gatto

See Also

See getMarkerClasses to get the classes only. See markers for details about spatial markers storage and encoding.

Examples

library("pRolocdata")
data(dunkley2006)
## marker vectors
myVmarkers <- getMarkers(dunkley2006)
head(myVmarkers)
## marker matrix
dunkley2006 <- mrkVecToMat(dunkley2006, mfcol = "Markers")
myMmarkers <- getMarkers(dunkley2006, fcol = "Markers")
head(myMmarkers)

Description

This function computes and outputs normalised distances from a "ClustDistList" object.

Usage

getNormDist(object, p = 1/3)

Arguments

Value

An numeric of normalised distances, one per protein set in the ClustDistList.

Author(s)

Lisa Breckels

See Also

"ClustDistList", "ClustDist", and examples in clustDist.

Description

Convenience accessor to the predicted feature localisation in an 'MSnSet'. This function returns the predictions of an MSnSet instance. As a side effect, it prints out a prediction table.

Usage

getPredictions(object, fcol, scol, mcol = "markers", t = 0, verbose = TRUE)

Arguments

Value

An instance of class "MSnSet" with fcol.pred feature variable storing the prediction results according to the chosen threshold.

Author(s)

Laurent Gatto and Lisa Breckels

See Also

orgQuants for calculating organelle-specific thresholds.

Examples

library("pRolocdata")
data(dunkley2006)
res <- svmClassification(dunkley2006, fcol = "pd.markers",
                         sigma = 0.1, cost = 0.5)
fData(res)$svm[500:510]
fData(res)$svm.scores[500:510]
getPredictions(res, fcol = "svm", t = 0) ## all predictions
getPredictions(res, fcol = "svm", t = .9) ## single threshold 
## 50% top predictions per class
ts <- orgQuants(res, fcol = "svm", t = .5)
getPredictions(res, fcol = "svm", t = ts)

Description

Converts GO identifiers to/from GO terms, either explicitly or by checking if (any items in) the input contains "GO:".

Usage

goIdToTerm(x, names = TRUE, keepNA = TRUE)

goTermToId(x, names = TRUE, keepNA = TRUE)

flipGoTermId(x, names = TRUE, keepNA = TRUE)

prettyGoTermId(x)

Arguments

Value

A character of GO terms (ids) if x were ids (terms).

Author(s)

Laurent Gatto

Examples

goIdToTerm("GO:0000001")
goIdToTerm("GO:0000001", names = FALSE)
goIdToTerm(c("GO:0000001", "novalid"))
goIdToTerm(c("GO:0000001", "GO:0000002", "notvalid"))
goTermToId("mitochondrion inheritance")
goTermToId("mitochondrion inheritance", name = FALSE)
goTermToId(c("mitochondrion inheritance", "notvalid"))
prettyGoTermId("mitochondrion inheritance")
prettyGoTermId("GO:0000001")
flipGoTermId("mitochondrion inheritance")
flipGoTermId("GO:0000001")
flipGoTermId("GO:0000001", names = FALSE)

Description

Highlights a set of features of interest given as a FeaturesOfInterest instance on a PCA plot produced by plot2D or plot3D. If none of the features of interest are found in the MSnset's featureNames, an warning is thrown.

Usage

highlightOnPlot(object, foi, labels, args = list(), ...)

highlightOnPlot3D(object, foi, labels, args = list(), radius = 0.1 * 3, ...)

Arguments

Value

NULL; used for its side effects.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data("tan2009r1")
x <- FeaturesOfInterest(description = "A test set of features of interest",
                        fnames = featureNames(tan2009r1)[1:10],
                        object = tan2009r1)

## using FeaturesOfInterest or feature names
par(mfrow = c(2, 1))
plot2D(tan2009r1)
highlightOnPlot(tan2009r1, x)
plot2D(tan2009r1)
highlightOnPlot(tan2009r1, featureNames(tan2009r1)[1:10])

.pca <- plot2D(tan2009r1)
head(.pca)
highlightOnPlot(.pca, x, col = "red")
highlightOnPlot(tan2009r1, x, col = "red", cex = 1.5)
highlightOnPlot(tan2009r1, x, labels = TRUE)

.pca <- plot2D(tan2009r1, dims = c(1, 3))
highlightOnPlot(.pca, x, pch = "+", dims = c(1, 3))
highlightOnPlot(tan2009r1, x, args = list(dims = c(1, 3)))

.pca2 <- plot2D(tan2009r1, mirrorX = TRUE, dims = c(1, 3))
## previous pca matrix, need to mirror X axis
highlightOnPlot(.pca, x, pch = "+", args = list(mirrorX = TRUE))
## new pca matrix, with X mirrors (and 1st and 3rd PCs)
highlightOnPlot(.pca2, x, col = "red")

plot2D(tan2009r1)
highlightOnPlot(tan2009r1, x)
highlightOnPlot(tan2009r1, x, labels = TRUE, pos = 3)
highlightOnPlot(tan2009r1, x, labels = "Flybase.Symbol", pos = 1)

## in 3 dimensions
if (interactive()) {
  plot3D(tan2009r1, radius1 = 0.05)
  highlightOnPlot3D(tan2009r1, x, labels = TRUE)
  highlightOnPlot3D(tan2009r1, x)
}

Description

Classification using for the k-nearest neighbours algorithm.

Usage

knnClassification(
  object,
  assessRes,
  scores = c("prediction", "all", "none"),
  k,
  fcol = "markers",
  ...
)

Arguments

Value

An instance of class "MSnSet" with knn and knn.scores feature variables storing the classification results and scores respectively.

Author(s)

Laurent Gatto

Examples

library(pRolocdata)
data(dunkley2006)
## reducing parameter search space and iterations 
params <- knnOptimisation(dunkley2006, k = c(3, 10), times = 3)
params
plot(params)
f1Count(params)
levelPlot(params)
getParams(params)
res <- knnClassification(dunkley2006, params)
getPredictions(res, fcol = "knn")
getPredictions(res, fcol = "knn", t = 0.75)
plot2D(res, fcol = "knn")

Description

Classification parameter optimisation for the k-nearest neighbours algorithm.

Usage

knnOptimisation(
  object,
  fcol = "markers",
  k = seq(3, 15, 2),
  times = 100,
  test.size = 0.2,
  xval = 5,
  fun = mean,
  seed,
  verbose = TRUE,
  ...
)

Arguments

Details

Note that when performance scores precision, recall and (macro) F1 are calculated, any NA values are replaced by 0. This decision is motivated by the fact that any class that would have either a NA precision or recall would result in an NA F1 score and, eventually, a NA macro F1 (i.e. mean(F1)). Replacing NAs by 0s leads to F1 values of 0 and a reduced yet defined final macro F1 score.

Value

An instance of class "GenRegRes".

Author(s)

Laurent Gatto

See Also

knnClassification and example therein.

Description

Classification using a variation of the KNN implementation of Wu and Dietterich's transfer learning schema

Usage

knntlClassification(
  primary,
  auxiliary,
  fcol = "markers",
  bestTheta,
  k,
  scores = c("prediction", "all", "none"),
  seed
)

Arguments

Value

A character vector of the classifications for the unknowns

Author(s)

Lisa Breckels

See Also

knntlOptimisation

Examples

library(pRolocdata)
data(andy2011)
data(andy2011goCC)
## reducing calculation time of k by pre-running knnOptimisation
x <- c(andy2011, andy2011goCC)
k <- lapply(x, function(z)
            knnOptimisation(z, times=5,
                            fcol = "markers.orig",
                            verbose = FALSE))
k <- sapply(k, function(z) getParams(z))
k
## reducing parameter search with theta = 1,
## weights of only 1 or 0 will be considered
opt <- knntlOptimisation(andy2011, andy2011goCC,
                         fcol = "markers.orig",
                         times = 2,
                         by = 1, k = k)
opt
th <- getParams(opt)
plot(opt)
res <- knntlClassification(andy2011, andy2011goCC,
                           fcol = "markers.orig", th, k)
res

Description

Classification parameter optimisation for the KNN implementation of Wu and Dietterich's transfer learning schema

Usage

knntlOptimisation(
  primary,
  auxiliary,
  fcol = "markers",
  k,
  times = 50,
  test.size = 0.2,
  xval = 5,
  by = 0.5,
  length.out,
  th,
  xfolds,
  BPPARAM = BiocParallel::bpparam(),
  method = "Breckels",
  log = FALSE,
  seed
)

Arguments

Details

knntlOptimisation implements a variation of Wu and Dietterich's transfer learning schema: P. Wu and T. G. Dietterich. Improving SVM accuracy by training on auxiliary data sources. In Proceedings of the Twenty-First International Conference on Machine Learning, pages 871 - 878. Morgan Kaufmann, 2004. A grid search for the best theta is performed.

Value

A list of containing the theta combinations tested, associated macro F1 score and accuracy for each combination over each round (specified by times).

Author(s)

Lisa Breckels

References

Breckels LM, Holden S, Wonjar D, Mulvey CM, Christoforou A, Groen AJ, Kohlbacher O, Lilley KS, Gatto L. Learning from heterogeneous data sources: an application in spatial proteomics. bioRxiv. doi: http://dx.doi.org/10.1101/022152

Wu P, Dietterich TG. Improving SVM Accuracy by Training on Auxiliary Data Sources. Proceedings of the 21st International Conference on Machine Learning (ICML); 2004.

See Also

knntlClassification and example therein.

Description

Classification using the support vector machine algorithm.

Usage

ksvmClassification(
  object,
  assessRes,
  scores = c("prediction", "all", "none"),
  cost,
  fcol = "markers",
  ...
)

Arguments

Value

An instance of class "MSnSet" with ksvm and ksvm.scores feature variables storing the classification results and scores respectively.

Author(s)

Laurent Gatto

Examples

library(pRolocdata)
data(dunkley2006)
## reducing parameter search space and iterations 
params <- ksvmOptimisation(dunkley2006, cost = 2^seq(-1,4,5), times = 3)
params
plot(params)
f1Count(params)
levelPlot(params)
getParams(params)
res <- ksvmClassification(dunkley2006, params)
getPredictions(res, fcol = "ksvm")
getPredictions(res, fcol = "ksvm", t = 0.75)
plot2D(res, fcol = "ksvm")

Description

Classification parameter optimisation for the support vector machine algorithm.

Usage

ksvmOptimisation(
  object,
  fcol = "markers",
  cost = 2^(-4:4),
  times = 100,
  test.size = 0.2,
  xval = 5,
  fun = mean,
  seed,
  verbose = TRUE,
  ...
)

Arguments

Details

Note that when performance scores precision, recall and (macro) F1 are calculated, any NA values are replaced by 0. This decision is motivated by the fact that any class that would have either a NA precision or recall would result in an NA F1 score and, eventually, a NA macro F1 (i.e. mean(F1)). Replacing NAs by 0s leads to F1 values of 0 and a reduced yet defined final macro F1 score.

Value

An instance of class "GenRegRes".

Author(s)

Laurent Gatto

See Also

ksvmClassification and example therein.

Description

Creates a new "MSnSet" instance populated with a GO term binary matrix based on an original object.

Usage

makeGoSet(object, params, namespace = "cellular_component", evidence = NULL)

Arguments

Value

A new "MSnSet" with the GO terms for the respective features in the original object.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(dunkley2006)
data(dunkley2006params)
goset <- makeGoSet(dunkley2006[1:10, ],
                   dunkley2006params)
goset
exprs(goset)
image(goset)

Description

These functions implement the T augmented Gaussian mixture (TAGM) model for mass spectrometry-based spatial proteomics datasets using the maximum a posteriori (MAP) optimisation routine.

Usage

## S4 method for signature 'MAPParams'
show(object)

logPosteriors(x)

tagmMapTrain(
  object,
  fcol = "markers",
  method = "MAP",
  numIter = 100,
  mu0 = NULL,
  lambda0 = 0.01,
  nu0 = NULL,
  S0 = NULL,
  beta0 = NULL,
  u = 2,
  v = 10,
  seed = NULL
)

tagmMapPredict(
  object,
  params,
  fcol = "markers",
  probJoint = FALSE,
  probOutlier = TRUE
)

Arguments

Details

The tagmMapTrain function generates the MAP parameters (object or class MAPParams) based on an annotated quantitative spatial proteomics dataset (object of class MSnbase::MSnSet). Both are then passed to the tagmPredict function to predict the sub-cellular localisation of protein of unknown localisation. See the pRoloc-bayesian vignette for details and examples. In this implementation, if numerical instability is detected in the covariance matrix of the data a small multiple of the identity is added. A message is printed if this conditioning step is performed.

Value

tagmMapTrain returns an instance of class MAPParams().

tagmPredict returns an instance of class MSnbase::MSnSet containing the localisation predictions as a new tagm.map.allocation feature variable.

Slots

method

A character() storing the TAGM method name.

priors

A list() with the priors for the parameters

seed

An integer() with the random number generation seed.

posteriors

A list() with the updated posterior parameters and log-posterior of the model.

datasize

A list() with details about size of data

Author(s)

Laurent Gatto

Oliver M. Crook

References

A Bayesian Mixture Modelling Approach For Spatial Proteomics Oliver M Crook, Claire M Mulvey, Paul D. W. Kirk, Kathryn S Lilley, Laurent Gatto bioRxiv 282269; doi: https://doi.org/10.1101/282269

See Also

The plotEllipse() function can be used to visualise TAGM models on PCA plots with ellipses. The tagmMapTrain() function to use the TAGM MAP method.

Description

These function extract the marker or unknown proteins into a new MSnSet.

Usage

markerMSnSet(object, fcol = "markers")

unknownMSnSet(object, fcol = "markers")

Arguments

Value

An new MSnSet with marker/unknown proteins only.

Author(s)

Laurent Gatto

See Also

sampleMSnSet testMSnSet and markers for markers encoding.

Examples

library("pRolocdata")
data(dunkley2006)
mrk <- markerMSnSet(dunkley2006)
unk <- unknownMSnSet(dunkley2006)
dim(dunkley2006)
dim(mrk)
dim(unk)
table(fData(dunkley2006)$markers)
table(fData(mrk)$markers)
table(fData(unk)$markers)
## matrix-encoded markers
dunkley2006 <- mrkVecToMat(dunkley2006)
dim(markerMSnSet(dunkley2006, "Markers"))
stopifnot(all.equal(featureNames(markerMSnSet(dunkley2006, "Markers")),
                    featureNames(markerMSnSet(dunkley2006, "markers"))))
dim(unknownMSnSet(dunkley2006, "Markers"))
stopifnot(all.equal(featureNames(unknownMSnSet(dunkley2006, "Markers")),
                    featureNames(unknownMSnSet(dunkley2006, "markers"))))

Description

Internal infrastructure to query/handle several individual mart instance. See MartInterface.R for details.

Author(s)

Laurent Gatto <[email protected]>

Description

Helper function to get the number of outlier at each MCMC iteration.

Helper function to get mean component allocation at each MCMC iteration.

Helper function to get mean probability of belonging to outlier at each iteration.

Wrapper for the geweke diagnostics from coda package also return p-values.

Helper function to pool chains together after processing

Helper function to burn n iterations from the front of the chains

Helper function to subsample the chains, known informally as thinning.

Produces a violin plot with the protein posterior probabilities distributions for all organelles.

Usage

mcmc_get_outliers(x)

mcmc_get_meanComponent(x)

mcmc_get_meanoutliersProb(x)

geweke_test(k)

mcmc_pool_chains(param)

mcmc_burn_chains(x, n = 50)

mcmc_thin_chains(x, freq = 5)

## S4 method for signature 'MCMCParams,character'
plot(x, y, ...)

Arguments

Value

A list of length length(x).

A list of length length(x).

A list of length length(x).

A matrix with the test z- and p-values for each chain.

A pooled MCMCParams object.

An updated MCMCParams object.

A thinned 'MCMCParams' object.

A ggplot2 object.

Author(s)

Laurent Gatto

Description

The MCMCParams infrastructure is used to store and process Marchov chain Monte Carlo results for the T-Augmented Gaussian Mixture model (TAGM) from Crook et al. (2018).

Usage

chains(object)

## S4 method for signature 'MCMCParams'
show(object)

## S4 method for signature 'ComponentParam'
show(object)

## S4 method for signature 'MCMCChain'
show(object)

## S4 method for signature 'MCMCChains'
length(x)

## S4 method for signature 'MCMCParams'
length(x)

## S4 method for signature 'MCMCChains,ANY,ANY'
x[[i, j = "missing", drop = "missing"]]

## S4 method for signature 'MCMCParams,ANY,ANY'
x[[i, j = "missing", drop = "missing"]]

## S4 method for signature 'MCMCChains,ANY,ANY,ANY'
x[i, j = "missing", drop = "missing"]

## S4 method for signature 'MCMCParams,ANY,ANY,ANY'
x[i, j = "missing", drop = "missing"]

## S4 method for signature 'MCMCChains'
show(object)

Arguments

Details

Objects of the MCMCParams class are created with the tagmMcmcTrain() function. These objects store the priors of the generative TAGM model and the results of the MCMC chains, which themselves are stored as an instance of class MCMCChains and can be accessed with the chains() function. A summary of the MCMC chains (or class MCMCSummary) can be further computed with the tagmMcmcProcess() function.

See the pRoloc-bayesian vignette for examples.

Slots

chains

list() containing the individual full MCMC chain results in an MCMCChains instance. Each element must be a valid MCMCChain instance.

posteriorEstimates

A data.frame documenting the prosterior priors in an MCMCSummary instance. It contains N rows and columns tagm.allocation, tagm.probability, tagm.outlier, tagm.probability.lowerquantile, tagm.probability.upperquantile and tagm.mean.shannon.

diagnostics

A matrix of dimensions 1 by 2 containing the MCMCSummary diagnostics.

tagm.joint

A matrix of dimensions N by K storing the joint probability in an MCMCSummary instance.

method

character(1) describing the method in the MCMCParams object.

chains

Object of class MCMCChains containing the full MCMC chain results stored in the MCMCParams object.

priors

list()

summary

Object of class MCMCSummary the summarised MCMC results available in the MCMCParams instance.

n

integer(1) indicating the number of MCMC interactions. Stored in an MCMCChain instance.

K

integer(1) indicating the number of components. Stored in an MCMCChain instance.

N

integer(1) indicating the number of proteins. Stored in an MCMCChain instance.

Component

matrix(N, n) component allocation results of an MCMCChain instance.

ComponentProb

matrix(N, n, K) component allocation probabilities of an MCMCChain instance.

Outlier

matrix(N, n) outlier allocation results.

OutlierProb

matrix(N, n, 2) outlier allocation probabilities of an MCMCChain instance.

See Also

The function tagmMcmcTrain() to construct object of this class.

Description

This function updates an MSnSet instances and sets markers class to unknown if there are less than n instances.

Usage

minMarkers(object, n = 10, fcol = "markers")

Arguments

Value

An instance of class "MSnSet" with a new feature variables, named after the original fcol variable and the n value.

Author(s)

Laurent Gatto

See Also

getPredictions to filter based on classification scores.

Examples

library(pRolocdata)
data(dunkley2006)
d2 <- minMarkers(dunkley2006, 20)
getMarkers(dunkley2006)
getMarkers(d2, fcol = "markers20")

Description

Model calibration model with posterior z-scores and posterior shrinkage

Usage

mixing_posterior_check(object, params, priors, fcol = "markers")

Arguments

Value

Used for side effect of producing plot. Invisibily returns an ggplot object that can be further manipulated

Author(s)

Oliver M. Crook <[email protected]>

Examples

## Not run: 
library("pRoloc")
data("tan2009r1")

tanres <- tagmMcmcTrain(object = tan2009r1)
tanres <- tagmMcmcProcess(tanres)
tan2009r1 <- tagmMcmcPredict(object = tan2009r1, params = tanres, probJoint = TRUE)
myparams <- chains(e14Tagm_converged_pooled)[[1]]
myparams2 <- chains(mcmc_pool_chains(tanres))[[1]]
priors <- tanres@priors
pRoloc:::mixing_posterior_check(object = tan2009r1, params = myparams2, priors = priors)

## End(Not run)

Description

This method implements MLInterfaces' MLean method for instances of the class "MSnSet".

Methods

signature(formula = "formula", data = "MSnSet", .method = "learnerSchema", trainInd = "numeric")

The learning problem is stated with the formula and applies the .method schema on the MSnSet data input using the trainInd numeric indices as train data.

signature(formula = "formula", data = "MSnSet", .method = "learnerSchema", trainInd = "xvalSpec")

In this case, an instance of xvalSpec is used for cross-validation.

signature(formula = "formula", data = "MSnSet", .method = "clusteringSchema", trainInd = "missing")

Hierarchical (hclustI), k-means (kmeansI) and partitioning around medoids (pamI) clustering algorithms using MLInterface's MLearn interface.

See Also

The MLInterfaces package documentation, in particular MLearn.

Description

Given two MSnSet instances of one MSnSetList with at least two items, this function produces an animation that shows the transition from the first data to the second.

Usage

move2Ds(object, pcol, fcol = "markers", n = 25, hl)

Arguments

Value

Used for its side effect of producing a short animation.

Author(s)

Laurent Gatto

See Also

plot2Ds to a single figure with the two datasets.

Examples

library("pRolocdata")
data(dunkley2006)

## Create a relevant MSnSetList using the dunkley2006 data
xx <- split(dunkley2006, "replicate")
xx1 <- xx[[1]]
xx2 <- xx[[2]]
fData(xx1)$markers[374] <- "Golgi"
fData(xx2)$markers[412] <- "unknown"
xx@x[[1]] <- xx1
xx@x[[2]] <- xx2

## The features we want to track
foi <- FeaturesOfInterest(description = "test",
                          fnames = featureNames(xx[[1]])[c(374, 412)])

## (1) visualise each experiment separately
par(mfrow = c(2, 1))
plot2D(xx[[1]], main = "condition A")
highlightOnPlot(xx[[1]], foi)
plot2D(xx[[2]], mirrorY = TRUE, main = "condition B")
highlightOnPlot(xx[[2]], foi, args = list(mirrorY = TRUE))

## (2) plot both data on the same plot
par(mfrow = c(1, 1))
tmp <- plot2Ds(xx) 
highlightOnPlot(data1(tmp), foi, lwd = 2)
highlightOnPlot(data2(tmp), foi, pch = 5, lwd = 2)

## (3) create an animation
move2Ds(xx, pcol = "replicate")
move2Ds(xx, pcol = "replicate", hl = foi)

Description

A function to calculate average marker profiles.

Usage

mrkConsProfiles(object, fcol = "markers", method = mean)

Arguments

Value

A matrix of dimensions number of clusters (exluding unknowns) by number of fractions.

Author(s)

Laurent Gatto and Lisa M. Breckels

See Also

The mrkHClust function to produce a hierarchical cluster.

Examples

library("pRolocdata")
data(dunkley2006)
mrkConsProfiles(dunkley2006)
mrkConsProfiles(dunkley2006, method = median)
mm <- mrkConsProfiles(dunkley2006)
## Reorder fractions
o <- order(dunkley2006$fraction)
## Plot mean organelle profiles using the
## default pRoloc colour palette.
matplot(t(mm[, o]), type = "l",
        xlab = "Fractions", ylab = "Relative intensity",
        main = "Mean organelle profiles",
        col = getStockcol(), lwd = 2, lty = 1)
## Add a legend
addLegend(markerMSnSet(dunkley2006), where = "topleft")

Description

This functions calculates an average protein profile for each marker class (proteins of unknown localisation are ignored) and then generates a dendrogram representing the relation between marker classes. The colours used for the dendrogram labels are taken from the default colours (see getStockcol) so as to match the colours with other spatial proteomics visualisations such as plot2D.

Usage

mrkHClust(
  object,
  fcol = "markers",
  distargs,
  hclustargs,
  method = mean,
  plot = TRUE,
  ...
)

Arguments

Value

Invisibly returns a dendrogram object, containing the hierarchical cluster as computed by hclust.

Author(s)

Laurent Gatto

Examples

library("pRolocdata")
data(dunkley2006)
mrkHClust(dunkley2006)

Description

Functions producing a new vector (matrix) marker vector set from an existing matrix (vector) marker set.

Usage

mrkVecToMat(object, vfcol = "markers", mfcol = "Markers")

mrkMatToVec(object, mfcol = "Markers", vfcol = "markers")

mrkMatAndVec(object, vfcol = "markers", mfcol = "Markers")

showMrkMat(object, mfcol = "Markers")

isMrkMat(object, fcol = "Markers")

isMrkVec(object, fcol = "markers")

mrkEncoding(object, fcol = "markers")

Arguments

Details

Sub-cellular markers can be encoded in two different ways. Sets of spatial markers can be represented as character vectors (character or factor, to be accurate), stored as feature metadata, and proteins of unknown or uncertain localisation (unlabelled, to be classified) are marked with the "unknown" character. While very handy, this encoding suffers from some drawbacks, in particular the difficulty to label proteins that reside in multiple (possible or actual) localisations. The markers vector feature data is typically named markers. A new matrix encoding is also supported. Each spatial compartment is defined in a column in a binary markers matrix and the resident proteins are encoded with 1s. The markers matrix feature data is typically named Markers. If proteins are assigned unique localisations only (i.e. no multi-localisation) or their localisation is unknown (unlabelled), then both encodings are equivalent. When the markers are encoded as vectors, features of unknown localisation are defined as fData(object)[, fcol] == "unknown". For matrix-encoded markers, unlabelled proteins are defined as rowSums(fData(object)[, fcol]) == 0.

The mrkMatToVec and mrkVecToMat functions enable the conversion from matrix (vector) to vector (matrix). The mrkMatAndVec function generates the missing encoding from the existing one. If the destination encoding already exists, or, more accurately, if the feature variable of the destination encoding exists, an error is thrown. During the conversion from matrix to vector, if multiple possible label exists, they are dropped, i.e. they are converted to "unknown". Function isMrkVec and isMrkMat can be used to test if a marker set is encoded as a vector or a matrix. mrkEncoding returns either "vector" or "matrix" depending on the nature of the markers.

Value

An updated MSnSet with a new vector (matrix) marker set.

Author(s)

Laurent Gatto and Lisa Breckels

See Also

Other functions that operate on markers are getMarkers, getMarkerClasses and markerMSnSet. To add markers to an existing MSnSet, see the addMarkers function and pRolocmarkers, for a list of suggested markers.

Examples

library("pRolocdata")
data(dunkley2006)
dunk <- mrkVecToMat(dunkley2006)
head(fData(dunk)$Markers)
fData(dunk)$markers <- NULL
dunk <- mrkMatToVec(dunk)
stopifnot(all.equal(fData(dunkley2006)$markers,
                    fData(dunk)$markers))

Description

Classification using the naive Bayes algorithm.

Usage

nbClassification(
  object,
  assessRes,
  scores = c("prediction", "all", "none"),
  laplace,
  fcol = "markers",
  ...
)

Arguments

Value

An instance of class "MSnSet" with nb and nb.scores feature variables storing the classification results and scores respectively.

Author(s)

Laurent Gatto

Examples

library(pRolocdata)
data(dunkley2006)
## reducing parameter search space and iterations 
params <- nbOptimisation(dunkley2006, laplace = c(0, 5),  times = 3)
params
plot(params)
f1Count(params)
levelPlot(params)
getParams(params)
res <- nbClassification(dunkley2006, params)
getPredictions(res, fcol = "naiveBayes")
getPredictions(res, fcol = "naiveBayes", t = 1)
plot2D(res, fcol = "naiveBayes")

Description

Classification algorithm parameter for the naive Bayes algorithm.

Usage

nbOptimisation(
  object,
  fcol = "markers",
  laplace = seq(0, 5, 0.5),
  times = 100,
  test.size = 0.2,
  xval = 5,
  fun = mean,
  seed,
  verbose = TRUE,
  ...
)

Arguments

Details

Note that when performance scores precision, recall and (macro) F1 are calculated, any NA values are replaced by 0. This decision is motivated by the fact that any class that would have either a NA precision or recall would result in an NA F1 score and, eventually, a NA macro F1 (i.e. mean(F1)). Replacing NAs by 0s leads to F1 values of 0 and a reduced yet defined final macro F1 score.

Value

An instance of class "GenRegRes".

Author(s)

Laurent Gatto

See Also

nbClassification and example therein.

Description

Produces a pca plot with uncertainty in organelle means projected onto the PCA plot with contours.

Usage

nicheMeans2D(
  object,
  params,
  priors,
  dims = c(1, 2),
  fcol = "markers",
  aspect = 0.5
)

Arguments

Value

Used for side effect of producing plot. Invisibily returns an ggplot object that can be further manipulated

Author(s)

Oliver M. Crook <[email protected]>

Examples

## Not run: 
library("pRolocdata")
data("tan2009r1")

tanres <- tagmMcmcTrain(object = tan2009r1)
tanres <- tagmMcmcProcess(tanres)
tan2009r1 <- tagmMcmcPredict(object = tan2009r1, params = tanres, probJoint = TRUE)
myparams <- chains(e14Tagm_converged_pooled)[[1]]
myparams2 <- chains(mcmc_pool_chains(tanres))[[1]]
priors <- tanres@priors
pRoloc:::nicheMeans2D(object = tan2009r1, params = myparams2, priors = priors)

## End(Not run)

Description

Methods computing the nearest neighbour indices and distances for matrix and MSnSet instances.

Methods

signature(object = "matrix", k = "numeric", dist = "character", ...)

Calculates indices and distances to the k (default is 3) nearest neighbours of each feature (row) in the input matrix object. The distance dist can be either of "euclidean" or "mahalanobis". Additional parameters can be passed to the internal function FNN::get.knn. Output is a matrix with 2 * k columns and nrow(object) rows.

signature(object = "MSnSet", k = "numeric", dist = "character", ...)

As above, but for an MSnSet input. The indices and distances to the k nearest neighbours are added to the object's feature metadata.

signature(object = "matrix", query = "matrix", k = "numeric", ...)

If two matrix instances are provided as input, the k (default is 3) indices and distances of the nearest neighbours of query in object are returned as a matrix of dimensions 2 * k by nrow(query). Additional parameters are passed to FNN::get.knnx. Only euclidean distance is available.

Examples

library("pRolocdata")
data(dunkley2006)

## Using a matrix as input
m <- exprs(dunkley2006)
m[1:4, 1:3]
head(nndist(m, k = 5))
tail(nndist(m[1:100, ], k = 2, dist = "mahalanobis"))

## Same as above for MSnSet
d <- nndist(dunkley2006, k = 5)
head(fData(d))

d <- nndist(dunkley2006[1:100, ], k = 2, dist = "mahalanobis")
tail(fData(d))

## Using a query
nndist(m[1:100, ], m[101:110, ], k = 2)

Description

Classification using the artificial neural network algorithm.

Usage

nnetClassification(
  object,
  assessRes,
  scores = c("prediction", "all", "none"),
  decay,
  size,
  fcol = "markers",
  ...
)

Arguments

Value

An instance of class "MSnSet" with nnet and nnet.scores feature variables storing the classification results and scores respectively.

Author(s)

Laurent Gatto

Examples

library(pRolocdata)
data(dunkley2006)
## reducing parameter search space and iterations 
params <- nnetOptimisation(dunkley2006, decay = 10^(c(-1, -5)), size = c(5, 10), times = 3)
params
plot(params)
f1Count(params)
levelPlot(params)
getParams(params)
res <- nnetClassification(dunkley2006, params)
getPredictions(res, fcol = "nnet")
getPredictions(res, fcol = "nnet", t = 0.75)
plot2D(res, fcol = "nnet")

Description

Classification parameter optimisation for artificial neural network algorithm.

Usage

nnetOptimisation(
  object,
  fcol = "markers",
  decay = c(0, 10^(-1:-5)),
  size = seq(1, 10, 2),
  times = 100,
  test.size = 0.2,
  xval = 5,
  fun = mean,
  seed,
  verbose = TRUE,
  ...
)

Arguments

Details

Note that when performance scores precision, recall and (macro) F1 are calculated, any NA values are replaced by 0. This decision is motivated by the fact that any class that would have either a NA precision or recall would result in an NA F1 score and, eventually, a NA macro F1 (i.e. mean(F1)). Replacing NAs by 0s leads to F1 values of 0 and a reduced yet defined final macro F1 score.

Value

An instance of class "GenRegRes".

Author(s)

Laurent Gatto

See Also

nnetClassification and example therein.

Description

For a given matrix of annotation information, this function returns the information ordered according to the best fit with the data.

Usage

orderGoAnnotations(
  object,
  fcol = "GOAnnotations",
  k = 1:5,
  n = 5,
  p = 1/3,
  verbose = TRUE,
  seed
)

Arguments

Details

As there are typically many protein/annotation sets that may fit the data we order protein sets by best fit i.e. cluster tightness, by computing the mean normalised Euclidean distance for all instances per protein set.

For each protein set i.e. proteins that have been labelled with a specified term/information criteria, we find the best k cluster components for the set (the default is to testk = 1:5) according to the minimum mean normalised pairwise Euclidean distance over all component clusters. (Note: when testing k if any components are found to have less than n proteins these components are not included and k is reduced by 1).

Each component cluster is normalised by N^p (where N is the total number of proteins per component, and p is the power). Hueristally, p = 1/3 and normalising by N^1/3 has been found the optimum normalisation factor.

Candidates in the matrix are ordered according to lowest mean normalised pairwise Euclidean distance as we expect high density, tight clusters to have the smallest mean normalised distance.

This function is a wrapper for running clustDist, getNormDist, see the "Annotating spatial proteomics data" vignette for more details.

Value

An updated MSnSet containing the newly ordered fcol matrix.

Author(s)

Lisa M Breckels

See Also

addGoAnnotations and example therein.

Description

This function produces organelle-specific quantiles corresponding to the given classification scores.

Usage

orgQuants(object, fcol, scol, mcol = "markers", t, verbose = TRUE)

Arguments

Value

A named vector of organelle thresholds.

Author(s)

Lisa Breckels

See Also

getPredictions to get organelle predictions based on calculated thresholds.

Examples

library("pRolocdata")
data(dunkley2006)
res <- svmClassification(dunkley2006, fcol = "pd.markers",
                         sigma = 0.1, cost = 0.5)
## 50% top predictions per class
ts <- orgQuants(res, fcol = "svm", t = .5)
getPredictions(res, fcol = "svm", t = ts)

Description

Classification using the PerTurbo algorithm.

Usage

perTurboClassification(
  object,
  assessRes,
  scores = c("prediction", "all", "none"),
  pRegul,
  sigma,
  inv,
  reg,
  fcol = "markers"
)

Arguments

Value

An instance of class "MSnSet" with perTurbo and perTurbo.scores feature variables storing the classification results and scores respectively.

Author(s)

Thomas Burger and Samuel Wieczorek

References

N. Courty, T. Burger, J. Laurent. "PerTurbo: a new classification algorithm based on the spectrum perturbations of the Laplace-Beltrami operator", The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2011), D. Gunopulos et al. (Eds.): ECML PKDD 2011, Part I, LNAI 6911, pp. 359 - 374, Athens, Greece, September 2011.

Examples

library(pRolocdata)
data(dunkley2006)
## reducing parameter search space 
params <- perTurboOptimisation(dunkley2006,
                               pRegul = 2^seq(-2,2,2),
                               sigma = 10^seq(-1, 1, 1),
                               inv = "Inversion Cholesky",
                               reg ="tikhonov",
                               times = 3)
params
plot(params)
f1Count(params)
levelPlot(params)
getParams(params)
res <- perTurboClassification(dunkley2006, params)
getPredictions(res, fcol = "perTurbo")
getPredictions(res, fcol = "perTurbo", t = 0.75)
plot2D(res, fcol = "perTurbo")

Description

Classification parameter optimisation for the PerTurbo algorithm

Usage

perTurboOptimisation(
  object,
  fcol = "markers",
  pRegul = 10^(seq(from = -1, to = 0, by = 0.2)),
  sigma = 10^(seq(from = -1, to = 1, by = 0.5)),
  inv = c("Inversion Cholesky", "Moore Penrose", "solve", "svd"),
  reg = c("tikhonov", "none", "trunc"),
  times = 1,
  test.size = 0.2,
  xval = 5,
  fun = mean,
  seed,
  verbose = TRUE
)

Arguments

Details

Note that when performance scores precision, recall and (macro) F1 are calculated, any NA values are replaced by 0. This decision is motivated by the fact that any class that would have either a NA precision or recall would result in an NA F1 score and, eventually, a NA macro F1 (i.e. mean(F1)). Replacing NAs by 0s leads to F1 values of 0 and a reduced yet defined final macro F1 score.

Value

An instance of class "GenRegRes".

Author(s)

Thomas Burger and Samuel Wieczorek

See Also

perTurboClassification and example therein.

Description

phenoDisco is a semi-supervised iterative approach to detect new protein clusters.

Usage

phenoDisco(
  object,
  fcol = "markers",
  times = 100,
  GS = 10,
  allIter = FALSE,
  p = 0.05,
  ndims = 2,
  modelNames = mclust.options("emModelNames"),
  G = 1:9,
  BPPARAM,
  tmpfile,
  seed,
  verbose = TRUE,
  dimred = c("PCA", "t-SNE"),
  ...
)

Arguments

Details

The algorithm performs a phenotype discovery analysis as described in Breckels et al. Using this approach one can identify putative subcellular groupings in organelle proteomics experiments for more comprehensive validation in an unbiased fashion. The method is based on the work of Yin et al. and used iterated rounds of Gaussian Mixture Modelling using the Expectation Maximisation algorithm combined with a non-parametric outlier detection test to identify new phenotype clusters.

One requires 2 or more classes to be labelled in the data and at a very minimum of 6 markers per class to run the algorithm. The function will check and remove features with missing values using the filterNA method.

A parallel implementation, relying on the BiocParallel package, has been added in version 1.3.9. See the BPPARAM arguent for details.

Important: Prior to version 1.1.2 the row order in the output was different from the row order in the input. This has now been fixed and row ordering is now the same in both input and output objects.

Value

An instance of class MSnSet containing the phenoDisco predictions.

Author(s)

Lisa M. Breckels <[email protected]>

References

Yin Z, Zhou X, Bakal C, Li F, Sun Y, Perrimon N, Wong ST. Using iterative cluster merging with improved gap statistics to perform online phenotype discovery in the context of high-throughput RNAi screens. BMC Bioinformatics. 2008 Jun 5;9:264. PubMed PMID: 18534020.

Breckels LM, Gatto L, Christoforou A, Groen AJ, Lilley KS and Trotter MWB. The Effect of Organelle Discovery upon Sub-Cellular Protein Localisation. J Proteomics. 2013 Aug 2;88:129-40. doi: 10.1016/j.jprot.2013.02.019. Epub 2013 Mar 21. PubMed PMID: 23523639.

Examples

## Not run: 
library(pRolocdata)
data(tan2009r1)
pdres <- phenoDisco(tan2009r1, fcol = "PLSDA")
getPredictions(pdres, fcol = "pd", scol = NULL)
plot2D(pdres, fcol = "pd")

## to pre-process the data with t-SNE instead of PCA
pdres <- phenoDisco(tan2009r1, fcol = "PLSDA", dimred = "t-SNE")

## End(Not run)

Description

Generate 2 or 3 dimensional feature distribution plots to illustrate localistation clusters. Rows/features containing NA values are removed prior to dimension reduction except for the "nipals" method. For this method, it is advised to set the method argument 'ncomp' to a low number of dimensions to avoid computing all components when analysing large datasets.

Usage

plot2D(
  object,
  fcol = "markers",
  fpch,
  unknown = "unknown",
  dims = 1:2,
  score = 1,
  method = "PCA",
  methargs,
  axsSwitch = FALSE,
  mirrorX = FALSE,
  mirrorY = FALSE,
  col,
  pch,
  cex,
  index = FALSE,
  idx.cex = 0.75,
  addLegend,
  identify = FALSE,
  plot = TRUE,
  grid = TRUE,
  ...
)

## S4 method for signature 'MSnSet'
plot3D(
  object,
  fcol = "markers",
  dims = c(1, 2, 3),
  radius1 = 0.1,
  radius2 = radius1 * 2,
  plot = TRUE,
  ...
)

Arguments

Details

plot3D relies on the ##' rgl package, that will be loaded automatically.

• Note that plot2D has been update in version 1.3.6 to support more organelle classes than colours defined in getStockcol. In such cases, the default colours are recycled using the default plotting characters defined in getStockpch. See the example for an illustration. The alpha argument is also depreciated in version 1.3.6. Use setStockcol to set colours with transparency instead. See example below.

• Version 1.11.3: to plot data as is, i.e. without any transformation, method can be set to "none" (as opposed to passing pre-computed values to method as a matrix, in previous versions). If object is an MSnSet, the untransformed values in the assay data will be plotted. If object is a matrix with coordinates, then a matching MSnSet must be passed to methargs.

Value

Used for its side effects of generating a plot. Invisibly returns the 2 or 3 dimensions that are plotted.

Author(s)

Laurent Gatto <[email protected]>

See Also

addLegend to add a legend to plot2D figures (the legend is added by default on plot3D) and plotDist for alternative graphical representation of quantitative organelle proteomics data. plot2Ds to overlay 2 data sets on the same PCA plot. The plotEllipse function can be used to visualise TAGM models on PCA plots with ellipses.

Examples

library("pRolocdata")
data(dunkley2006)
plot2D(dunkley2006, fcol = NULL)
plot2D(dunkley2006, fcol = NULL, col = "black")
plot2D(dunkley2006, fcol = "markers")
addLegend(dunkley2006,
          fcol = "markers",
          where = "topright",
          cex = 0.5, bty = "n", ncol = 3)
title(main = "plot2D example")
## available methods
plot2Dmethods
plot2D(dunkley2006, fcol = NULL, method = "kpca", col = "black")
plot2D(dunkley2006, fcol = NULL, method = "kpca", col = "black",
       methargs = list(kpar = list(sigma = 1)))
plot2D(dunkley2006, method = "lda")
plot2D(dunkley2006, method = "hexbin")
## Using transparent colours
setStockcol(paste0(getStockcol(), "80"))
plot2D(dunkley2006, fcol = "markers")
## New behavious in 1.3.6 when not enough colours
setStockcol(c("blue", "red", "green"))
getStockcol() ## only 3 colours to be recycled
getMarkers(dunkley2006)
plot2D(dunkley2006)
## reset colours
setStockcol(NULL)
plot2D(dunkley2006, method = "none") ## plotting along 2 first fractions
plot2D(dunkley2006, dims = c(3, 5), method = "none") ## plotting along fractions 3 and 5
## pre-calculate PC1 and PC2 coordinates
pca <- plot2D(dunkley2006, plot=FALSE)
head(pca)
plot2D(pca, method = "none", methargs  = list(dunkley2006))

## plotting in 3 dimenstions
plot3D(dunkley2006)
plot3D(dunkley2006, radius2 = 0.3)
plot3D(dunkley2006, dims = c(2, 4, 6))