msImpute: Function for imputing missing values in data.

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

Description

Replaces missing values with non-zero estimates calculated using a selected method.

Usage

msImpute(
  data,
  imputeMethod = c("halfmin", "bpca", "knn", "rf"),
  kKnn = 5,
  nPcs = 3,
  maxIterRf = 10,
  nTreeRf = 100,
  compoundsAsNeighbors = FALSE,
  compVars = c("mz", "rt"),
  sampleVars = c("subject_id"),
  colExtraText = NULL,
  separator = NULL,
  missingValue = NA,
  returnToSE = FALSE,
  returnToDF = FALSE
)

Arguments

data	Data set as either a data frame or 'SummarizedExperiement'.
imputeMethod	String specifying imputation method. Options are "halfmin" (half the minimum value), "bpca" (Bayesian PCA), and "knn" (k-nearest neighbors).
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.
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'.
returnToSE	Logical value indicating whether to return as 'SummarizedExperiment'
returnToDF	Logical value indicating whether to return as data frame.

Value

A data frame or 'SummarizedExperiment' with missing data imputed. Default return type is set to match the data input but may be altered with the 'returnToSE' or 'returnToDF' arguments.

References

Oba, S.et al.(2003) A Bayesian missing value estimation for gene expression profile data. Bioinformatics, 19, 2088-2096

Stacklies, W.et al.(2007) pcaMethods A bioconductor package providing PCA methods for incomplete data. Bioinformatics, 23, 1164-1167.

A. Kowarik, M. Templ (2016) Imputation with R package VIM. Journal of Statistical Software, 74(7), 1-16.

Stekhoven D. J., & Buehlmann, P. (2012). MissForest - non-parametric missing value imputation for mixed-type data. Bioinformatics, 28(1), 112-118.

Examples

# Load, tidy, summarize, and filter example dataset
data(msquant)

summarizedDF <- msSummarize(msquant,
                            compVars = c("mz", "rt"),
                            sampleVars = c("spike", "batch", "replicate", 
                            "subject_id"),
                            cvMax = 0.50,
                            minPropPresent = 1/3,
                            colExtraText = "Neutral_Operator_Dif_Pos_",
                            separator = "_",
                            missingValue = 1)
                            
filteredDF <- msFilter(summarizedDF,
                       filterPercent = 0.8,
                       compVars = c("mz", "rt"),
                       sampleVars = c("spike", "batch", "subject_id"),
                       separator = "_")

                           
# Impute dataset using 3 possible options
hmImputedDF <- msImpute(filteredDF, imputeMethod = "halfmin",
                        compVars = c("mz", "rt"),
                        sampleVars = c("spike", "batch", "subject_id"),
                        separator = "_",
                        missingValue = 0)

bpcaImputedDF <- msImpute(filteredDF, imputeMethod = "bpca",
                          nPcs = 3,
                          compVars = c("mz", "rt"),
                          sampleVars = c("spike", "batch", "subject_id"),
                          separator = "_",
                          missingValue = 0)

knnImputedDF <- msImpute(filteredDF, imputeMethod = "knn",
                         kKnn = 5,
                         compVars = c("mz", "rt"),
                         sampleVars = c("spike", "batch", "subject_id"),
                         separator = "_",
                         missingValue = 0)                                

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