Mercurial > repos > computational-metabolomics > mspurity_puritya
view frag4feature.R @ 11:416460f611cc draft default tip
planemo upload for repository https://github.com/computational-metabolomics/mspurity-galaxy commit 94f746247c464fb270358421399810d8a9fb9e8a
author | computational-metabolomics |
---|---|
date | Fri, 13 Sep 2024 13:42:16 +0000 |
parents | 10e56e53668c |
children |
line wrap: on
line source
library(optparse) library(msPurity) library(xcms) print(sessionInfo()) xset_pa_filename_fix <- function(opt, pa, xset = NULL) { if (!is.null(opt$mzML_files) && !is.null(opt$galaxy_names)) { # NOTE: Relies on the pa@fileList having the names of files given as 'names' of the variables # needs to be done due to Galaxy moving the files around and screwing up any links to files filepaths <- trimws(strsplit(opt$mzML_files, ",")[[1]]) # nolint filepaths <- filepaths[filepaths != ""] galaxy_names <- trimws(strsplit(opt$galaxy_names, ",")[[1]]) galaxy_names <- galaxy_names[galaxy_names != ""] nsave <- names(pa@fileList) old_filenames <- basename(pa@fileList) pa@fileList <- filepaths[match(names(pa@fileList), galaxy_names)] names(pa@fileList) <- nsave pa@puritydf$filename <- basename(pa@fileList[match(pa@puritydf$filename, old_filenames)]) pa@grped_df$filename <- basename(pa@fileList[match(pa@grped_df$filename, old_filenames)]) } print(pa@fileList) if (!is.null(xset)) { print(xset@filepaths) if (!all(basename(pa@fileList) == basename(xset@filepaths))) { if (!all(names(pa@fileList) == basename(xset@filepaths))) { print("FILELISTS DO NOT MATCH") message("FILELISTS DO NOT MATCH") quit(status = 1) } else { xset@filepaths <- unname(pa@fileList) } } } return(list(pa, xset)) } option_list <- list( make_option(c("-o", "--out_dir"), type = "character"), make_option("--pa", type = "character"), make_option("--xset", type = "character"), make_option("--ppm", default = 10), make_option("--plim", default = 0.0), make_option("--convert2RawRT", action = "store_true"), make_option("--intense", action = "store_true"), make_option("--createDB", action = "store_true"), make_option("--cores", default = 4), make_option("--mzML_files", type = "character"), make_option("--galaxy_names", type = "character"), make_option("--grp_peaklist", type = "character"), make_option("--useGroup", action = "store_true") ) # store options opt <- parse_args(OptionParser(option_list = option_list)) print(opt) loadRData <- function(rdata_path, name) { # loads an RData file, and returns the named xset object if it is there load(rdata_path) return(get(ls()[ls() %in% name])) } # This function retrieve a xset like object # @author Gildas Le Corguille lecorguille@sb-roscoff.fr getxcmsSetObject <- function(xobject) { # XCMS 1.x if (class(xobject) == "xcmsSet") { return(xobject) } # XCMS 3.x if (class(xobject) == "XCMSnExp") { # Get the legacy xcmsSet object suppressWarnings(xset <- as(xobject, "xcmsSet")) if (!is.null(xset@phenoData$sample_group)) { xcms::sampclass(xset) <- xset@phenoData$sample_group } else { xcms::sampclass(xset) <- "." } return(xset) } } # Requires pa <- loadRData(opt$pa, "pa") xset <- loadRData(opt$xset, c("xset", "xdata")) xset <- getxcmsSetObject(xset) pa@cores <- opt$cores print(pa@fileList) print(xset@filepaths) if (is.null(opt$intense)) { intense <- FALSE } else { intense <- TRUE } if (is.null(opt$convert2RawRT)) { convert2RawRT <- FALSE } else { convert2RawRT <- TRUE } if (is.null(opt$createDB)) { createDB <- FALSE } else { createDB <- TRUE } if (is.null(opt$useGroup)) { fix <- xset_pa_filename_fix(opt, pa, xset) pa <- fix[[1]] xset <- fix[[2]] useGroup <- FALSE } else { # if are only aligning to the group not eah file we do not need to align the files between the xset and pa object print("useGroup") fix <- xset_pa_filename_fix(opt, pa) pa <- fix[[1]] useGroup <- TRUE } if (is.null(opt$grp_peaklist)) { grp_peaklist <- NA } else { grp_peaklist <- opt$grp_peaklist } print(useGroup) pa <- msPurity::frag4feature( pa = pa, xset = xset, ppm = opt$ppm, plim = opt$plim, intense = intense, convert2RawRT = convert2RawRT, db_name = "alldata.sqlite", out_dir = opt$out_dir, grp_peaklist = grp_peaklist, create_db = createDB, use_group = useGroup ) print(pa) save(pa, file = file.path(opt$out_dir, "frag4feature_output.RData")) pa@grped_df$filename <- sapply(pa@grped_df$fileid, function(x) names(pa@fileList)[as.integer(x)]) print(head(pa@grped_df)) write.table(pa@grped_df, file.path(opt$out_dir, "frag4feature_output.tsv"), row.names = FALSE, sep = "\t")