Package 'RNAmodR.ML'

ModifierMLkeras class

Description

The ModifierMLkeras class extends the virtual class ModifierMLModel and unifies the access to Keras machine learning models used in the detection of post-transcriptional modifications in RNA sequencing data. The ModifierMLkeras class is virtual itself and must be extended from for each individual machine learning model.

The ModifierMLkeras class extends the virtual class ModifierMLModel and unifies the access to Keras machine learning models used in the detection of post-transcriptional modifications in RNA sequencing data. The ModifierMLkeras class is virtual itself and must be extended from for each individual machine learning model.

Since a stored model needs to be loaded from file, the additional slot modelFile is used and can be accessed through a function of the same name. Upon creation of a ModifierMLkeras object, the model is loaded from file, if modelFile is not an empty character value and a valid file name.

Usage

## S4 method for signature 'ModifierMLkeras,ModifierML'
useModel(x, y)

Arguments

x	a ModifierMLkeras object
y	a ModifierML object

Value

a ModifierMLkeras object

Slots

modelFile

a character vector of length == 1L, which describes a model to load via load_model_hdf5. The model is then stored in the model slot.

See Also

ModifierMLModel

---

ModifierMLranger class

Description

The ModifierMLranger class extends the virtual class ModifierMLModel and unifies the access to ranger machine learning models used in the detection of post-transcriptional modifications in RNA sequencing data. The ModifierMLranger class is virtual itself and must be extended from for each individual machine learning model.

Usage

## S4 method for signature 'ModifierMLranger,ModifierML'
useModel(x, y)

Arguments

x	a ModifierMLranger object
y	a ModifierML object

Value

a ModifierMLranger object

See Also

ModifierMLModel

Examples

# example class derived from the virtual ModifierMLranger class
setClass("ModifierMLexample",
         contains = c("ModifierMLranger"))
ModifierMLexample <- function(...){
  new("ModifierMLexample")
}
mlmodel <- ModifierMLexample()

---

RNAmodR.ML

Description

To extend the functionality of the RNAmodR package and classical detection strategies towards detection through machine learning models, RNAmodR.ML provides classes and an example workflow.

Author(s)

Felix G M Ernst [aut]

See Also

RNAmodR package.

---

Example data in the RNAmodR.ML package

Description

This contains an example ModifierSet object

Usage

data(dmod)

data(mod7)

data(me)

data(model)

Format

a GRanges object containg D positions

a GRanges object containg m7G positions

a ModMLExample object for examples

a ranger object for examples

---

RNAmodR.ML functions for example

Description

The exported functions here are used in the vignette as examples. If you want to reuse them, please implement them yourself. This will allow for fine tuning on your side and does not create a depency to example functions, which could change

Usage

aggregate_example(x)

find_mod_example(x, minCoverage)

Arguments

x	a ModifierML object
minCoverage	the minimum coverage for finding modifications

Value

• aggregate_example a SplitDataFrameList object containing the aggregated data for a Modifier object

• find_mod_example a GRanges object with coordinates of found modifications

Examples

# no examples runable. See vignette for explanation on how to use these
# example functions
aggregate_example
find_mod_example

---

The ModifierML class

Description

The ModifierML class is a virtual class, which provides the central functionality for searching with a machine learning models for patterns of post-transcriptional RNA modifications in high throughput sequencing data.

It extends the virtual Modifier class form the RNAmodR package and add the useMLModel function. If not called directly for a ModifierML class, the useMLModel will be called from the aggregate function.

The slot mlModel is added and serves a dual purpose. If mlModel is a character, a class of the type ModifierMLModel is created upon creation of a ModifierML object. However, for developing purposes the slot can also remain empty and a ModifierMLModel object can be set using the setMLModel function and retrieved using the getMLModel. If the mlModel slot is empty, the findMod setting will be set to FALSE and the ModifierML object will be returned just with the aggregate data. Such an object can then be used to train a machine learning model. The data can be accessed using trainingData.

Usage

setMLModel(x) <- value

getMLModel(x)

hasMLModel(x)

useMLModel(x)

## S4 replacement method for signature 'ModifierML'
setMLModel(x) <- value

## S4 method for signature 'ModifierML'
getMLModel(x)

## S4 method for signature 'ModifierML'
hasMLModel(x)

## S4 method for signature 'ModifierML'
aggregate(x, force = FALSE)

## S4 method for signature 'ModifierML'
useMLModel(x)

## S4 method for signature 'ModifierML'
modify(x, force = FALSE)

Arguments

x	a ModifierML object.
value	a ModifierMLModel object
force	whether to recreate the aggregated data, if it is already stored inside the Modifier object.

Value

a ModifierML object

Slots

mlModel

a character describing a class name for creating a ModifierMLModel object or a ModifierMLModel object itself. If mlModel is a character, the class will tried to be create by calling a function of the same name.

See Also

Modifier

Examples

# an example implementation of a ModifierML object
setClass("ModMLExample",
         contains = c("RNAModifierML"),
         prototype = list(mod = c("D"),
                          score = "score",
                          dataType = c("PileupSequenceData",
                                       "CoverageSequenceData"),
                          mlModel = character(0)))
# constructor function for ModMLExample
ModMLExample <- function(x, annotation = NA, sequences = NA, seqinfo = NA,
                         ...){
  RNAmodR:::Modifier("ModMLExample", x = x, annotation = annotation,
                     sequences = sequences, seqinfo = seqinfo, ...)
}

---

Assemble training data from aggregate sequence data

Description

trainingData subsets the aggregate data of a ModifierML object to positions as defined by coord. Positions with an entry in the mod column are labeled TRUE.

Usage

trainingData(x, coord, ...)

## S4 method for signature 'ModifierML,GRanges'
trainingData(x, coord, ...)

## S4 method for signature 'ModifierML,GRangesList'
trainingData(x, coord, ...)

Arguments

x	a ModifierML object
coord	a GRanges or a GRangesList object
...	See subsetByCoord for more details. type is hard coded to FALSE to disregard subsetting by type. In addition, merge is set to FALSE by default, but it can be set to TRUE.

Value

a CompressedSplitDataFrameList with aggregate data and an addition label column.

See Also

For more details have a look at subsetByCoord.

Examples

data("dmod",package = "RNAmodR.ML")
setClass("ModMLExample",
         contains = c("RNAModifierML"),
         prototype = list(mod = c("D"),
                          score = "score",
                          dataType = c("PileupSequenceData",
                                       "CoverageSequenceData"),
                          mlModel = character(0)))
data("me",package = "RNAmodR.ML")
nextUPos <- function(gr){
  nextU <- lapply(seq.int(1L,2L),
                  function(i){
                    subseq <- subseq(RNAmodR::sequences(me)[dmod$Parent], start(dmod)+3L)
                    pos <- start(dmod) + 2L +
                      vapply(strsplit(as.character(subseq),""),
                             function(y){which(y == "U")[i]},integer(1))
                    ans <- dmod
                    ranges(ans) <- IRanges(start = pos, width = 1L)
                    ans
                  })
  nextU <- do.call(c,nextU)
  nextU$mod <- NULL
  unique(nextU)
}
nondmod <- nextUPos(dmod)
nondmod <- nondmod[!(nondmod %in% dmod)]
coord <- unique(c(dmod,nondmod))
coord <- coord[order(as.integer(coord$Parent))]
trainingData(me,coord)

---

ModifierMLModel virtual class

Description

The ModifierMLModel is a virtual class and is used for representing different types of machine learning models used in the detection of post transcriptional modifiations in RNA sequencing data.

The next class inheriting from here should only implement a certain type of model and also be virtual. The grand child of the ModifierMLModel class should than implement a specific model for detecting certain types of modifications.

Usage

useModel(x, y)

## S4 method for signature 'ModifierMLModel,ANY'
useModel(x, y)

Arguments

x	a ModifierMLModel object
y	See ModifierMLranger for an example

Value

a ModifierMLModel object

Slots

model

a machine learning object of any type

See Also

ModifierMLranger ModifierMLkeras

ModifierMLranger for an example implementation

Examples

# an example implementation of a ModifierMLModel object using the
# ModifierMLModel derived class ModifierMLranger
setClass("ModifierMLexample",
         contains = c("ModifierMLranger"))
ModifierMLexample <- function(...){
  new("ModifierMLexample")
}

---