msPrepare: Summarize, filter, impute, transform and normalize...

In KechrisLab/MSPrep: Package for Summarizing, Filtering, Imputing, and Normalizing Metabolomics Data

Description

Wrapper function for the entire MSPrep pre-analytics pipeline. Calls msSummarize(), msFilter, msImpute(), and msNormalize().

Usage

msPrepare(
  data,
  cvMax = 0.5,
  minPropPresent = 1/3,
  filterPercent = 0.8,
  imputeMethod = c("halfmin", "bpca", "knn", "rf", "none"),
  kKnn = 5,
  nPcs = 3,
  maxIterRf = 10,
  nTreeRf = 100,
  compoundsAsNeighbors = FALSE,
  normalizeMethod = c("median", "ComBat", "quantile", "quantile + ComBat",
    "median + ComBat", "CRMN", "RUV", "SVA", "none"),
  nControl = 10,
  controls = NULL,
  nComp = 2,
  kRUV = 3,
  covariatesOfInterest = NULL,
  batch = NULL,
  transform = c("log10", "log2", "none"),
  replicate = "replicate",
  compVars = c("mz", "rt"),
  sampleVars = c("subject_id"),
  colExtraText = NULL,
  separator = NULL,
  missingValue = NA,
  returnSummaryDetails = FALSE,
  returnToSE = FALSE,
  returnToDF = FALSE
)

Arguments

data	Data set as either a data frame or 'SummarizedExperiement'.
cvMax	Decimal value from 0 to 1 representing the acceptable level of coefficient of variation between replicates.
minPropPresent	Decimal value from 0 to 1 representing the minimum proportion present to summarize with median or mean. Below this the compound will be set to 0.
filterPercent	Decimal value indicating filtration threshold. Compounds which are present in fewer samples than the specified proportion will be removed.
imputeMethod	String specifying imputation method. Options are "halfmin" (half the minimum value), "bpca" (Bayesian PCA), and "knn" (k-nearest neighbors), or "none" to skip imputation.
kKnn	Number of clusters for 'knn' method.
nPcs	Number of principle components used for re-estimation for 'bpca' method.
maxIterRf	Maximum number of iterations to be performed given the stopping criterion is not met beforehand for 'rf' method.
nTreeRf	Number of trees to grow in each forest for 'rf' method.
compoundsAsNeighbors	For KNN imputation. If TRUE, compounds will be used as neighbors rather than samples. Note that using compounds as neighbors is significantly slower than using samples.
normalizeMethod	Name of normalization method. "ComBat" (only ComBat batch correction), "quantile" (only quantile normalization), "quantile + ComBat" (quantile with ComBat batch correction), "median" (only median normalization), "median + ComBat" (median with ComBat batch correction), "CRMN" (cross-contribution compensating multiple standard normalization), "RUV" (remove unwanted variation), "SVA" (surrogate variable analysis), or "none" to skip normalization.
nControl	Number of controls to estimate/utilize (for CRMN and RUV).
controls	Vector of control identifiers. Leave blank for data driven controls. Vector of column numbers from metafin dataset of that control (for CRMN and RUV).
nComp	Number of factors to use in CRMN algorithm.
kRUV	Number of factors to use in RUV algorithm.
covariatesOfInterest	Sample variables used as covariates in normalization algorithms (required for ComBat, CRMN, and SVA).
batch	Name of the sample variable identifying batch.
transform	Select transformation to apply to data prior to normalization. Options are "log10", "log2", and "none".
replicate	Name of sample variable specifying replicate. Must match an element in 'sampleVars' or a column in the column data of a 'SummarizedExperiment'.
compVars	Vector of the columns which identify compounds. If a 'SummarizedExperiment' is used for 'data', row variables will be used.
sampleVars	Vector of the ordered sample variables found in each sample column.
colExtraText	Any extra text to ignore at the beginning of the sample columns names. Unused for 'SummarizedExperiments'.
separator	Character or text separating each sample variable in sample columns. Unused for 'SummarizedExperiment'.
missingValue	Specifies the abundance value which indicates missing data. May be a numeric or 'NA'.
returnSummaryDetails	Logical value specifying whether to return details of replicate summarization.
returnToSE	Logical value specifying whether to return as 'SummarizedExperiment'
returnToDF	Logical value specifying whether to return as data frame.

Value

A data frame or 'SummarizedExperiment' with summarized technical replicates (if present), filtered compounds, missing values imputed, and transformed and normalized abundances. Default return type is set to match the data input but may be altered with the 'returnToSE' or 'returnToDF' arguments.

Examples

# Load example data
data(msquant)

# Call function to tidy, summarize, filter, impute, and normalize data
peparedData <- msPrepare(msquant, cvMax = 0.50, minPropPresent = 1/3,
                         filterPercent = 0.8, imputeMethod = "halfmin",
                         normalizeMethod = "quantile",
                         compVars = c("mz", "rt"),
                         sampleVars = c("spike", "batch", "replicate", 
                                        "subject_id"),
                         colExtraText = "Neutral_Operator_Dif_Pos_",
                         separator = "_", missingValue = 1, 
                         returnToSE = FALSE)

KechrisLab/MSPrep documentation built on Feb. 2, 2022, 2:43 a.m.