rRoma.R: Perform ROMA on a datasets
In Albluca/rRoma: An r implementation of ROMA

Perform ROMA on a datasets

rRoma.R(ExpressionMatrix, ModuleList, centerData = TRUE, ExpFilter = FALSE,
  UseWeigths = FALSE, DefaultWeight = 1, MinGenes = 10, MaxGenes = 1000,
  ApproxSamples = 5, nSamples = 100, OutGeneNumber = 5, Ncomp = 10,
  OutGeneSpace = NULL, FixedCenter = TRUE,
  GeneOutDetection = "L1OutExpOut", GeneOutThr = 5, GeneSelMode = "All",
  SampleFilter = TRUE, MoreInfo = FALSE, PlotData = FALSE, PCADims = 2,
  PCSignMode = "none", PCSignThr = NULL, UseParallel = FALSE,
  nCores = NULL, ClusType = "PSOCK", SamplingGeneWeights = NULL,
  FillNAMethod = list(), Grouping = NULL, FullSampleInfo = FALSE,
  GroupPCSign = FALSE, CorMethod = "pearson",
  PCAType = "DimensionsAreGenes", SuppressWarning = FALSE,
  ShowParallelPB = TRUE)

`ExpressionMatrix`	matrix, a numeric matrix containing the gene expression information. Columns indicate samples and rows indicated genes.
`ModuleList`	list, gene module list
`centerData`	logical, should the gene expression values be centered over the samples?
`ExpFilter`	logical, should the samples be filtered?
`UseWeigths`	logical, should the weigths be used for PCA calculation?
`DefaultWeight`	integer scalar, the default weigth to us if no weith is specified by the modile file and an algorithm requiring weigths is used
`MinGenes`	integer, the minimum number of genes reported by a module available in the expression matrix to process the module
`MaxGenes`	integer, the maximum number of genes reported by a module available in the expression matrix to process the module
`ApproxSamples`	integer between 0 and 100 the approximation parameter to reuse samples. This is the minimal percentage variation to reuse samples. For example 5, means that samples re recalculated only if the number of genes in the geneset has increased by at least 5%.
`nSamples`	integer, the number of randomized gene sampled (per module)
`OutGeneNumber`	scalar, number of median-absolute-deviations away from median required for the total number of genes expressed in a sample to be called an outlier
`Ncomp`	iteger, number of principal components used to filter samples in the gene expression space
`OutGeneSpace`	scalar, number of median-absolute-deviations away from median required for in a sample to be called an outlier in the gene expression space. If set to NULL, the gene space filtering will not be performed.
`FixedCenter`	logical, should PCA with fixed center be used?
`GeneOutDetection`	character scalar, the algorithm used to filter genes in a module. Possible values are 'L1OutVarPerc': percentage variation relative to the median variance explained supported by a leave one out approach 'L1OutVarDC': dendrogram clustering statistics on variance explained supported by a leave one out approach 'L1OutExpOut': number of median-absolute-deviations away from median explined variance 'L1OutSdMean': Number of standard deviations away from the mean The option "L1OutExpOut" requires the scater package to be installed.
`GeneOutThr`	scalar, threshold used by gene filtering algorithm in the modules. It can represent maximum size of filtered cluster ("L1OutVarDC"), minimal percentage variation (L1OutVarPerc) or the number of median-absolute-deviations away from median ("L1OutExpOut")
`GeneSelMode`	character scalar, mode used to sample genes: all available genes ("All") or genes not present in the module ("Others")
`SampleFilter`	logical, should outlier detection be applied to sampled data as well?
`MoreInfo`	logical, shuold detailed information on the computation by printed?
`PlotData`	logical, shuold debugging plots by produced ?
`PCADims`	integer, the number of PCA dimensions to compute. Should be >= 2. Note that, the value 1 is allowed, but is not advisable under normal circumstances. Larger values decrease the error in the estimation of the explained variance but increase the computation time.
`PCSignMode`	characrter scalar, the modality to use to determine the direction of the principal components. The following options are currentlhy available: 'none' (The direction is chosen at random) 'PreferActivation': the direction is chosen in such a way that the sum of the projection is positive 'UseAllWeights': as 'PreferActivation', but the projections are multiplied by the weigths, missing weights are set to DefaultWeight 'UseKnownWeights': as 'UseAllWeights', but missing weigth are set to 0 'CorrelateAllWeightsByGene': the direction is chosen in such a way to maximise the positive correlation between the expression of genes with a positive (negative) weights and the (reversed) PC projections, missing weights are set to DefaultWeight 'CorrelateKnownWeightsByGene': as 'CorrelateAllWeights', but missing weights are set to 0 'CorrelateAllWeightsBySample': the direction is chosen in such a way to maximise the positive correlation between the expression of genes and the PC corrected weigth (i.e., PC weigths are multiplied by gene weigths), missing weights are set to DefaultWeight 'CorrelateKnownWeightsBySample': as 'CorrelateAllWeightsBySample', but missing weights are set to 0 If 'CorrelateAllWeights', 'CorrelateKnownWeights', 'CorrelateAllWeightsBySample' or 'CorrelateKnownWeightsBySample' are used and GroupPCSign is TRUE, the correltions will be computed on the groups defined by Grouping.
`PCSignThr`	numeric scalar, a quantile threshold to limit the projections (or weights) to use, e.g., if equal to .9 only the 10% of genes with the largest projection (or weights) in absolute value will be considered.
`UseParallel`	boolean, shuold a parallel environment be used? Note that using a parallel environment will increase the memorey usage as a copy of the gene expression matrix is needed for each core
`nCores`	integer, the number of cores to use if UseParallel is TRUE. Set to NULL for auto-detection
`ClusType`	string, the cluster type to use. The default value ("PSOCK") should be available on most systems, unix-like environments also support the "FORK", which should be faster.
`SamplingGeneWeights`	named vector, numeric. Weigth so use when correcting the sign of the PC for sampled data.
`FillNAMethod`	names list, additional parameters to pass to the mice function
`Grouping`	named vector, the groups associated with the sample.
`FullSampleInfo`	boolean, should full PC information be computed and saved for all the randomised genesets?
`GroupPCSign`	boolean, should grouping information to be used to orient PCs?
`CorMethod`	character string indicating which correlation coefficient is to be used for orienting the principal components. Can be "pearson", "kendall", or "spearman".
`PCAType`	character string, the type of PCA to perform. It can be "DimensionsAreGenes" or "DimensionsAreSamples"
`SuppressWarning`	boolean, should warnings be displayed? This option well be ignored in non-interactive sessions.
`ShowParallelPB`	boolean, should the progress bas be displayed when using parallel processing. Note that the progress bar is diaplayed via the pbapply package. This may slow donwn the computation, expecially with FORK clusters.