The impute method performs data imputation on QFeatures and
SummarizedExperiment instance using a variety of methods.
Users should proceed with care when imputing data and take precautions to assure that the imputation produce valid results, in particular with naive imputations such as replacing missing values with 0.
See MsCoreUtils::impute_matrix() for details on
the different imputation methods available and strategies.
impute
# S4 method for SummarizedExperiment
impute(object, method, ...)
# S4 method for QFeatures
impute(object, method, ..., i, name = "imputedAssay")An object of class standardGeneric of length 1.
A SummarizedExperiment or QFeatures object with
missing values to be imputed.
character(1) defining the imputation method. See
imputeMethods() for available ones. See
MsCoreUtils::impute_matrix() for details.
Additional parameters passed to the inner imputation
function. See MsCoreUtils::impute_matrix() for details.
A logical(1) or a character(1) that defines which
element of the QFeatures instance to impute. It cannot be
missing and must be of length one.
A character(1) naming the new assay name. Default
is imputedAssay.
MsCoreUtils::imputeMethods()
#> [1] "bpca" "knn" "QRILC" "MLE" "MLE2" "MinDet" "MinProb"
#> [8] "min" "zero" "mixed" "nbavg" "with" "RF" "none"
data(se_na2)
## table of missing values along the rows (proteins)
table(rowData(se_na2)$nNA)
#>
#> 0 1 2 3 4 8 9 10
#> 301 247 91 13 2 23 10 2
## table of missing values along the columns (samples)
colData(se_na2)$nNA
#> [1] 34 45 56 39 47 52 49 61 41 42 55 45 51 43 57 53
## non-random missing values
notna <- which(!rowData(se_na2)$randna)
length(notna)
#> [1] 35
notna
#> [1] 6 20 79 88 130 187 227 231 238 264 275 317 324 363 373 382 409 437 445
#> [20] 453 456 474 484 485 492 514 516 546 568 580 594 631 648 664 671
impute(se_na2, method = "min")
#> class: SummarizedExperiment
#> dim: 689 16
#> metadata(3): MSnbaseFiles MSnbaseProcessing MSnbaseVersion
#> assays(1): ''
#> rownames(689): AT1G09210 AT1G21750 ... AT4G11150 AT4G39080
#> rowData names(2): nNA randna
#> colnames(16): M1F1A M1F4A ... M2F8B M2F11B
#> colData names(1): nNA
if (require("imputeLCMD")) {
impute(se_na2, method = "QRILC")
impute(se_na2, method = "MinDet")
}
#> Loading required package: imputeLCMD
#> Loading required package: tmvtnorm
#> Loading required package: mvtnorm
#> Loading required package: Matrix
#>
#> Attaching package: ‘Matrix’
#> The following object is masked from ‘package:S4Vectors’:
#>
#> expand
#> Loading required package: gmm
#> Loading required package: sandwich
#> Loading required package: norm
#> This package has some major limitations
#> (for example, it does not work reliably when
#> the number of variables exceeds 30),
#> and has been superseded by the norm2 package.
#> Loading required package: pcaMethods
#>
#> Attaching package: ‘pcaMethods’
#> The following object is masked from ‘package:stats’:
#>
#> loadings
#> Loading required package: impute
#> Imputing along margin 2 (samples/columns).
#> Imputing along margin 2 (samples/columns).
#> class: SummarizedExperiment
#> dim: 689 16
#> metadata(3): MSnbaseFiles MSnbaseProcessing MSnbaseVersion
#> assays(1): ''
#> rownames(689): AT1G09210 AT1G21750 ... AT4G11150 AT4G39080
#> rowData names(2): nNA randna
#> colnames(16): M1F1A M1F4A ... M2F8B M2F11B
#> colData names(1): nNA
if (require("norm"))
impute(se_na2, method = "MLE")
#> Imputing along margin 2 (samples/columns).
#> Warning: NAs introduced by coercion to integer range
#> Iterations of EM:
#> 1...2...
#> class: SummarizedExperiment
#> dim: 689 16
#> metadata(3): MSnbaseFiles MSnbaseProcessing MSnbaseVersion
#> assays(1): ''
#> rownames(689): AT1G09210 AT1G21750 ... AT4G11150 AT4G39080
#> rowData names(2): nNA randna
#> colnames(16): M1F1A M1F4A ... M2F8B M2F11B
#> colData names(1): nNA
impute(se_na2, method = "mixed",
randna = rowData(se_na2)$randna,
mar = "knn", mnar = "QRILC")
#> Imputing along margin 1 (features/rows).
#> Imputing along margin 1 (features/rows).
#> class: SummarizedExperiment
#> dim: 689 16
#> metadata(3): MSnbaseFiles MSnbaseProcessing MSnbaseVersion
#> assays(1): ''
#> rownames(689): AT1G09210 AT1G21750 ... AT4G11150 AT4G39080
#> rowData names(2): nNA randna
#> colnames(16): M1F1A M1F4A ... M2F8B M2F11B
#> colData names(1): nNA
## neighbour averaging
x <- se_na2[1:4, 1:6]
assay(x)[1, 1] <- NA ## min value
assay(x)[2, 3] <- NA ## average
assay(x)[3, 1:2] <- NA ## min value and average
## 4th row: no imputation
assay(x)
#> M1F1A M1F4A M1F7A M1F11A M1F2B M1F5B
#> AT1G09210 NA 0.275500 0.21600 0.18525 0.465667 0.199667
#> AT1G21750 0.332000 0.279667 NA 0.16600 0.451500 0.200375
#> AT1G51760 NA NA 0.16825 0.18825 0.459750 0.214500
#> AT1G56340 0.336733 NA NA NA 0.487167 0.201833
assay(impute(x, "nbavg"))
#> Assuming values are ordered.
#> Imputing along margin 1 (features/rows).
#> M1F1A M1F4A M1F7A M1F11A M1F2B M1F5B
#> AT1G09210 0.166000 0.275500 0.2160000 0.18525 0.465667 0.199667
#> AT1G21750 0.332000 0.279667 0.2228335 0.16600 0.451500 0.200375
#> AT1G51760 0.166000 0.167125 0.1682500 0.18825 0.459750 0.214500
#> AT1G56340 0.336733 NA NA NA 0.487167 0.201833