Impute: Dropout imputation using different methods
In beyergroup/ADImpute: Adaptive Dropout Imputer (ADImpute)

Description Usage Arguments Details Value See Also Examples

Impute performs dropout imputation on normalized data, based on the choice of imputation methods.

Impute(data, sce = NULL, do = 'Ensemble', write = FALSE,
outdir = getwd(), method.choice = NULL, scale = 1, pseudo.count = 1,
labels = NULL, cell.clusters = 2, drop_thre = NULL, type = 'count',
tr.length = ADImpute::transcript_length,
cores = BiocParallel::bpworkers(BPPARAM),
BPPARAM = BiocParallel::SnowParam(type = "SOCK"),
net.coef = ADImpute::network.coefficients, net.implementation = 'iteration',
bulk = NULL, true.zero.thr = NULL, prob.mat = NULL, ...)

`data`	matrix; raw counts (genes as rows and samples as columns)
`sce`	SingleCellExperiment; normalized counts and associated metadata.
`do`	character; choice of methods to be used for imputation. Currently supported methods are `'Baseline'`, `'DrImpute'`, `'Network'`, and `'Ensemble'`. Defaults to `'Ensemble'`. Not case-sensitive. Can include one or more methods. Non-supported methods will be ignored.
`write`	logical; write intermediary and imputed objects to files?
`outdir`	character; path to directory where output files are written. Defaults to working directory
`method.choice`	character; best performing method in training data for each gene
`scale`	integer; scaling factor to divide all expression levels by (defaults to 1)
`pseudo.count`	integer; pseudo-count to be added to expression levels to avoid log(0) (defaults to 1)
`labels`	character; vector specifying the cell type of each column of `data`
`cell.clusters`	integer; number of cell subpopulations
`drop_thre`	numeric; between 0 and 1 specifying the threshold to determine dropout values
`type`	A character specifying the type of values in the expression matrix. Can be 'count' or 'TPM'
`tr.length`	matrix with at least 2 columns: 'hgnc_symbol' and 'transcript_length'
`cores`	integer; number of cores used for paralell computation
`BPPARAM`	parallel back-end to be used during parallel computation. See `BiocParallelParam-class`.
`net.coef`	matrix; network coefficients. Please provide if you don't want to use ADImpute's network model. Must contain one first column 'O' acconting for the intercept of the model and otherwise be an adjacency matrix with hgnc_symbols in rows and columns. Doesn't have to be squared. See `ADImpute::demo_net` for a small example.
`net.implementation`	character; either 'iteration', for an iterative solution, or 'pseudoinv', to use Moore-Penrose pseudo-inversion as a solution. 'pseudoinv' is not advised for big data.
`bulk`	vector of reference bulk RNA-seq, if available (average across samples)
`true.zero.thr`	if set to NULL (default), no true zero estimation is performed. Set to numeric value between 0 and 1 for estimation. Value corresponds to the threshold used to determine true zeros: if the probability of dropout is lower than `true.zero.thr`, the imputed entries are set to zero.
`prob.mat`	matrix of the same size as data, filled with the dropout probabilities for each gene in each cell
`...`	additional parameters to pass to network-based imputation

Values that are 0 in data are imputed according to the best-performing methods indicated in method.choice. Currently supported methods are:

Baseline: imputation with average expression across all cells in the dataset. See ImputeBaseline.
Previously published approaches: DrImpute and SAVER.
Network: leverages information from a gene regulatory network to predicted expression of genes that are not quantified based on quantified interacting genes, in the same cell. See ImputeNetwork.
Ensemble: is based on results on a training subset of the data at hand, indicating which method best predicts the expression of each gene. These results are supplied via method.choice. Applies the imputation results of the best performing method to the zero entries of each gene.

If 'Ensemble' is included in do, method.choice has to be provided (use output from EvaluateMethods()). Impute can create a directory imputation containing the imputation results of all methods in do. If true.zero.thr is set, dropout probabilities are computed using scImpute's framework. Expression values with dropout probabilities below true.zero.thr will be set back to 0 if imputed, as they likely correspond to true biological zeros (genes not expressed in cell) rather than technical dropouts (genes expressed but not captured). If sce is set, imputed values by the different methods are added as new assays to sce. Each assay corresponds to one imputation method. If true.zero.thr is set, only the values after filtering for biological zeros will be added. This is different from the output if sce is not set, where the original values before filtering and the dropout probability matrix are returned.

if sce is not set: returns a list of imputation results (normalized, log-transformed) for all selected methods in do. If true.zero.thr is defined, returns a list of 3 elements: 1) a list, imputations, containing the direct imputation results from each method; 2) a list, zerofiltered, containing the results of imputation in imputations after setting biological zeros back to zero; 3) a matrix, dropoutprobabilities, containing the dropout probability matrix used to set biological zeros.
if sce is set: returns a SingleCellExperiment with new assays, each corresponding to one of the imputation methods applied. If true.zero.thr is defined, the assays will contain the results after imputation and setting biological zeros back to zero.

EvaluateMethods, ImputeBaseline, ImputeDrImpute, ImputeNetwork, ImputeSAVER

# Normalize demo data
norm_data <- NormalizeRPM(demo_data)
# Impute with particular method(s)
imputed_data <- Impute(do = 'Network', data = norm_data[,1:10],
net.coef = ADImpute::demo_net)
imputed_data <- Impute(do = 'Network', data = norm_data[,1:10],
net.implementation = 'pseudoinv', net.coef = ADImpute::demo_net)