R/complexCentricWorkflow.R

In hafenr/MACode: Functions and code snippets for my master thesis

#' Convert the output CC cprophet to a format that looks like it was output by
#' the PC workflow.
#' @export
convertToPCComplexFeatureFormat <- function(complex.features,
                                            decoy.complex.protein.assoc,
                                            corum.complex.protein.assoc) {

    complex.features <- data.table::copy(complex.features)
    complex.features[, complex_id := gsub('pep_', '', transition_group_id)]

    n_subunits <- sapply(strsplit(complex.features$aggr_Fragment_Annotation, ';'), length)
    complex.features[, n_subunits := n_subunits]
    subunit.counts <- corum.complex.protein.assoc[, length(protein_id), by=complex_id]
    setnames(subunit.counts, 'V1', 'n_subunits_annotated')
    complex.features <- merge(complex.features, subunit.counts, by='complex_id', all.x=TRUE)

    complex.features <-
          data.table(complex_id=complex.features$complex_id,
                     center_rt=complex.features$RT,
                     left_boundary_rt=complex.features$leftWidth,
                     right_boundary_rt=complex.features$rightWidth,
                     subunit_ids=gsub(';', ',', complex.features$aggr_Fragment_Annotation),
                     d_score=complex.features$d_score,
                     n_subunits=complex.features$n_subunits,
                     n_subunits_annotated=complex.features$n_subunits_annotated)

    # complex.features[is_decoy == TRUE, n_subunits_annotated := n_subunits]
    # # Now add the n_subunits_annotated information for decoy complexes as well!
    complex.features[, is_decoy := grepl('DECOY', complex_id)]

    # Split the data since only the decoy complex features have to be
    # changed.
    complex.features.target <- complex.features[is_decoy == FALSE]
    complex.features.decoy <- subset(complex.features[is_decoy == TRUE],
                                     select=-n_subunits_annotated)

    decoy.complex.protein.assoc <- data.table::copy(decoy.complex.protein.assoc)
    setnames(decoy.complex.protein.assoc, 'protein_id', 'complex_id')
    setnames(decoy.complex.protein.assoc, 'peptide_id', 'protein_id')

    decoy.complex.protein.assoc[, n_subunits_annotated := length(protein_id),
                                by=complex_id]
    # Simplify data.table s.t. it contains only complexes and their size info
    decoy.complex.size.info <-
        decoy.complex.protein.assoc[, list(complex_id), n_subunits_annotated]
    setkey(decoy.complex.size.info)
    decoy.complex.size.info <- unique(decoy.complex.size.info)

    complex.features.decoy <-
          merge(complex.features.decoy, decoy.complex.size.info, by='complex_id',
                all.x=TRUE)

    rbind(complex.features.target, complex.features.decoy)
    # complex.features
}

#' Create input data wor the CC workflow.
#' All this function essentially does is to take a long list peptide trace
#' data.table, produce the protein traces, annotate them with their complex
#' membership, rename all the columns such that they are compatible with
#' cprophet.
#' @export
createCCInputData <- function(peptide.traces.long,
                              out.file.name,
                              complex.protein.assoc) {
    protein.traces.long <- produceProteinTraces(peptide.traces.long)
    protein.traces.with.complex <-
        merge(protein.traces.long, corum.complex.protein.assoc, by='protein_id',
              allow.cartesian=T)

    protein.traces.wide <- longProtTracesToWide(protein.traces.with.complex)
    protein.traces.wide <- protein.traces.wide[, complex_name := NULL]

    setnames(protein.traces.wide, 'protein_id', 'peptide_id')
    setnames(protein.traces.wide, 'complex_id', 'protein_id')

    write.table(protein.traces.wide, out.file.name, sep='\t', row.names=FALSE)

}