applySeqGate: Performs a Filtering of Lowly Expressed Features
In SeqGate: Filtering of Lowly Expressed Features

Description Usage Arguments Details Value Author(s) References Examples

View source: R/seqgate.R

Implements the SeqGate methods to filter lowly expressed features (e.g. genes)

1 2	applySeqGate(readCountsSE, assayName, colCond, prop0=NULL, percentile=NULL, propUpThresh=0.9)

`readCountsSE`	`SummarizedExperiment` object including the `assayName` assay, that is a matrix of read counts. Rows correspond to features (genes or other genomic features) and columns correspond to sample libraries. The read counts must contain zeros, as the SeqGate method is based on the distribution of counts in replicates along with zero counts.
`assayName`	character string giving the name of the assay in `readCountsSE` for which we want to perform the filtering.
`colCond`	character string giving the name of the column describing the biological conditions to which the samples belong to, in the colData DataFrame of `readCountsSE`.
`prop0`	minimal proportion of zeros among a condition to consider that the feature is not or lowly expressed. `prop0` must be > 0 and < 1. See details for default setting.
`percentile`	percentile used on the 'max' distribution to determine the filtering threshold value. `percentile` must be >= 0 and <=1. See details for default setting.
`propUpThresh`	proportion of counts to be above the threshold in at least one condition to keep the feature. `propUpThresh` must be > 0 and <= 1. See details for default setting.

In order to find a theshold value to filter lowly expressed features, SeqGate analyzes the distribution of counts found in replicates along with zero counts. More specifically, features with a proportion of at least prop0 zeros in one condition are selected. The distribution of counts found in replicates of that same condition along with those zeros is computed. The chosen threshold is the count value corresponding to the percentile of this distribution. Finally, features having a proportion propUpThresh of replicates with counts below that value in all conditions are filtered.

If prop0 is not provided, it is set to the number of replicates minus one divided by the max total number of replicates. For example, if one of the condition of the experiment counts 2 replicates and the other condition 4 replicates, the proportion prop0 will be set to (4-1)/4=0.75.

If percentile is not provided, it is set to 0.90 if at least one of the conditions have 5 replicates or less. Otherwise, it is set to a value comprised between 0.5 and 0.9, depending on the number or replicates.

The input SummarizedExperiment with the following added elements:

`rowData(readCountsSE)$onFilter`	vector which `length = nrow(readCountsSE)` indicating if features are kept after filtering (`value = TRUE`) or not (`value = FALSE`)
`metadata(readCountsSE)$threshold`	applied filter threshold.

Christelle Reynès christelle.reynes@igf.cnrs.fr,
Stéphanie Rialle stephanie.rialle@mgx.cnrs.fr

Rialle, R. et al. (2020): SeqGate: a bioconductor package to perform data-driven filtering of RNAseq datasets (manuscript in preparation)

    # Loading of input data frame
    data(data_MiTF_1000genes)
    # Annotating conditions
    cond<-c("A","A","B","B","A","B")
    # Setting the SummarizedExperiment input
    rowData <- DataFrame(row.names=rownames(data_MiTF_1000genes))
    colData <- DataFrame(Conditions=cond)
    counts_strub <- SummarizedExperiment(
                        assays=list(counts=data_MiTF_1000genes),
                        rowData=rowData,
                        colData=colData)
    # Applying SeqGate
    counts_strub <- applySeqGate(counts_strub,"counts","Conditions")
    # Getting the matrix of kept genes after filtering
    keptGenes <- assay(counts_strub[rowData(counts_strub)$onFilter == TRUE,])