Mercurial > repos > mora-lab > chipenrich
changeset 0:63ec097240bf draft
Uploaded
author | mora-lab |
---|---|
date | Thu, 20 May 2021 08:41:45 +0000 |
parents | |
children | 3eaa000a7bf1 |
files | chipenrich.R |
diffstat | 1 files changed, 141 insertions(+), 0 deletions(-) [+] |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/chipenrich.R Thu May 20 08:41:45 2021 +0000 @@ -0,0 +1,141 @@ +############################################################################### +# title: chipenrich +# author: Xiaowei +# time: Mar.31 2021 +############################################################################### + +spec <- matrix(c("input_peaks", "P",1, "character", "peaks file", + "input_genome","G",1, "character", "genome", + "input_locusdef","L",1,"character", "locusdef", + "input_mappability","Map",0, "logical", "mappability", + "input_qc_plot","Q",0, "logical", "qc_plot", + "input_method","M",1,"character", "method", + "input_geneset","GS",1,"character", "geneset_category", + "input_minSize","Min",1,"integer","minSize", + "input_maxSize","Max",1,"integer", "maxSize", + "input_randomization","random",1,"character","randomization", + "input_reglocation","reg",1,"character","reglocation", + "input_num_peak_threshould","threshould",1,"numeric", "peak_threshould", + "output_peaks","out_peaks",1,"character","output_peaks", + "output_peaks_per_gene","out_ppg",1,"character","peaks per gene", + "output_enrich_results","out_er",1,"character", "enrich result", + "output_qc_plot", "out_qc", 1,"character","output_qc_plot" + ),byrow = TRUE, ncol = 5) + +opt <- getopt::getopt(spec) + + +#---------------------------------------------------------------- + +input_peaks <- read.csv(opt$input_peaks,header = TRUE) +input_genome = opt$input_genome +input_locusdef = opt$input_locusdef + +if (!is.null(opt$input_mappability)){ + input_mappability = 24 #根据表格来填写 +}else{ + input_mappability = NULL #supported_read_lengths() +} + +if(!is.null(opt$input_qc_plot)){ + input_qc_plot = opt$input_qc_plot + input_output_name = "x" +}else{ + input_qc_plot = FALSE + input_output_name = NULL +} + + +if (!is.null(opt$input_method)){input_method = opt$input_method}else{input_method ="chipenrich"} +if(!is.null(opt$input_geneset)){input_geneset = strsplit(opt$input_geneset, split = ",")[[1]]}else{input_geneset = NULL} +if(!is.null(opt$input_minSize)){input_minSize = opt$input_minSize}else{input_minSize = 15} +if(!is.null(opt$input_maxSize)){input_maxSize = opt$input_maxSize}else{input_maxSize = 2000} +if(!is.null(opt$input_randomization) & opt$input_randomization != "NULL"){input_randomization = opt$input_randomization}else{input_randomization = NULL} +input_core = parallel::detectCores() +if(!is.null(opt$input_reglocation)){input_reglocation = opt$input_reglocation}else{input_reglocation = "tss"} #proxReg +#chipenrich、hybridenrich +if(!is.null(opt$input_num_peak_threshould)){input_num_peak_threshould = opt$input_num_peak_threshould}else{input_num_peak_threshould = 1} +#================================================================ +#run codes +#================================================================ +suppressPackageStartupMessages(library(chipenrich)) +## ---- chipenrich---------------------------------------- +if (input_method == "chipenrich"){ + results = chipenrich(peaks = input_peaks, genome = input_genome, genesets = input_geneset, + locusdef = input_locusdef, qc_plots = FALSE, out_name = input_output_name, + mappability = input_mappability, + min_geneset_size = input_minSize, + max_geneset_size = input_maxSize, + randomization = input_randomization, + num_peak_threshold = input_num_peak_threshould, + n_cores = input_core) +} + + +## ----polyenrich---------------------------------------- +if (input_method == "polyenrich"){ + results = polyenrich(peaks = input_peaks, genome = input_genome, genesets = input_geneset, + method = 'polyenrich', + locusdef = input_locusdef, qc_plots = FALSE, out_name = input_output_name, + mappability = input_mappability, + min_geneset_size = input_minSize, + max_geneset_size = input_maxSize, + randomization = input_randomization, + n_cores = input_core) +} + +## ----hybridenrich---------------------------------------- +if (input_method == "hybridenrich"){ + results = hybridenrich(peaks = input_peaks, genome = input_genome, genesets = input_geneset, + locusdef = input_locusdef, qc_plots = FALSE, out_name = input_output_name, + mappability = input_mappability, + min_geneset_size = input_minSize, + max_geneset_size = input_maxSize, + randomization = input_randomization, + num_peak_threshold = input_num_peak_threshould, + n_cores = input_core) +} + +## ----proxReg---------------------------------------- +if (input_method == "proxReg"){ + results = proxReg(peaks = input_peaks, reglocation = input_reglocation, + genome = input_genome, genesets=input_geneset, + min_geneset_size = input_minSize, + max_geneset_size = input_maxSize, + randomization = input_randomization, + out_name=input_output_name, + n_cores = input_core) +} + +## ----broadenrich----------------------------------------peaks_H3K4me3_GM12878--- +if (input_method == "broadenrich"){ + results = broadenrich(peaks = input_peaks, genome = input_genome, genesets = input_geneset, + locusdef = input_locusdef, qc_plots = FALSE, out_name = input_output_name, + mappability = input_mappability, + min_geneset_size = input_minSize, + max_geneset_size = input_maxSize, + randomization = input_randomization, + n_cores = input_core) +} + +#=============================================================================== +# 输出 +#=============================================================================== +output_peaks = opt$output_peaks +output_peaks_per_gene = opt$output_peaks_per_gene +output_enrich_result = opt$output_enrich_results + +if (!is.null(output_peaks)){ + write.csv(results$peaks, file = output_peaks) +} + +if (!is.null(output_peaks_per_gene)){ + if (input_method != "proxReg"){ + write.csv(results$peaks_per_gene, file = output_peaks_per_gene) + } +} + +if (!is.null(output_enrich_result)){ + write.csv(results$results, file = output_enrich_result) +} +