annotate extract_putative_ltr.R @ 9:1aa578e6c8b3 draft

"planemo upload commit 9488b982bae902f1868785ec4ad47134dac50ff3"
author petr-novak
date Wed, 29 Jun 2022 09:25:54 +0000
parents 9de392f2fc02
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
1 #!/usr/bin/env Rscript
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
2 initial_options <- commandArgs(trailingOnly = FALSE)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
3 file_arg_name <- "--file="
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
4 script_name <- normalizePath(sub(file_arg_name, "", initial_options[grep(file_arg_name, initial_options)]))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
5 script_dir <- dirname(script_name)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
6 library(optparse)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
7
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
8 parser <- OptionParser()
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
9 option_list <- list(
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
10 make_option(c("-g", "--gff3"), action = "store", type = "character",
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
11 help = "gff3 with dante results", default = NULL),
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
12 make_option(c("-s", "--reference_sequence"), action = "store", type = "character",
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
13 help = "reference sequence as fasta", default = NULL),
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
14 make_option(c("-o", "--output"), action = "store", type = "character",
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
15 help = "output file path and prefix", default = NULL),
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
16 make_option(c("-c", "--cpu"), type = "integer", default = 5,
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
17 help = "Number of cpu to use [default %default]", metavar = "number"),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
18 make_option(c("-M", "--max_missing_domains"), type = "integer", default = 0,
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
19 help = "Maximum number of missing domains is retrotransposon [default %default]",
8
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
20 metavar = "number"),
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
21 make_option(c("-L", "--min_relative_length"), type = "numeric", default = 0.6,
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
22 help = "Minimum relative length of protein domain to be considered for retrostransposon detection [default %default]",
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
23 metavar = "number")
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
24
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
25 )
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
26 description <- paste(strwrap(""))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
27
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
28 epilogue <- ""
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
29 parser <- OptionParser(option_list = option_list, epilogue = epilogue, description = description,
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
30 usage = "usage: %prog COMMAND [OPTIONS]")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
31 opt <- parse_args(parser, args = commandArgs(TRUE))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
32
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
33
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
34 # load packages
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
35 suppressPackageStartupMessages({
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
36 library(rtracklayer)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
37 library(Biostrings)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
38 library(BSgenome)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
39 library(parallel)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
40 })
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
41
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
42
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
43 # CONFIGURATION
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
44 OFFSET <- 15000
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
45 # load configuration files and functions:
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
46 lineage_file <- paste0(script_dir, "/databases/lineage_domain_order.csv")
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
47 FDM_file <- paste0(script_dir, "/databases/feature_distances_model.RDS")
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
48 trna_db <- paste0(script_dir, "/databases/tRNAscan-SE_ALL_spliced-no_plus-old-tRNAs_UC_unique-3ends.fasta")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
49 ltr_utils_r <- paste0(script_dir, "/R/ltr_utils.R")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
50 if (file.exists(lineage_file) & file.exists(trna_db)) {
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
51 lineage_info <- read.table(lineage_file, sep = "\t", header = TRUE, as.is = TRUE)
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
52 FDM <- readRDS(FDM_file)
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
53 source(ltr_utils_r)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
54 }else {
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
55 # this destination work is installed using conda
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
56 lineage_file <- paste0(script_dir, "/../share/dante_ltr/databases/lineage_domain_order.csv")
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
57 FDM_file <- paste0(script_dir, "/../share/dante_ltr/databases/feature_distances_model.RDS")
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
58 trna_db <- paste0(script_dir, "/../share/dante_ltr/databases/tRNAscan-SE_ALL_spliced-no_plus-old-tRNAs_UC_unique-3ends.fasta")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
59 ltr_utils_r <- paste0(script_dir, "/../share/dante_ltr/R/ltr_utils.R")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
60 if (file.exists(lineage_file) & file.exists((trna_db))) {
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
61 lineage_info <- read.table(lineage_file, sep = "\t", header = TRUE, as.is = TRUE)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
62 source(ltr_utils_r)
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
63 FDM <- readRDS(FDM_file)
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
64 }else(
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
65 stop("configuration files not found")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
66 )
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
67 }
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
68
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
69
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
70 # for testing
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
71 if (FALSE) {
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
72 g <- rtracklayer::import("/mnt/raid/454_data/cuscuta/Ceuropea_assembly_v4/0_final_asm_hifiasm+longstitch/repeat_annotation/DANTE_on_CEUR_filtered_short_names.gff3")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
73 s <- readDNAStringSet("/mnt/raid/454_data/cuscuta/Ceuropea_assembly_v4/0_final_asm_hifiasm+longstitch/asm.bp.p.ctg_scaffolds.short_names.fa")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
74 lineage_info <- read.table("/mnt/raid/users/petr/workspace/ltr_finder_test/lineage_domain_order.csv", sep = "\t", header = TRUE, as.is = TRUE)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
75
8
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
76 g <- rtracklayer::import("/mnt/raid/users/petr/workspace/dante_ltr/test_data/sample_DANTE_unfiltered.gff3")
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
77 g <- rtracklayer::import("/mnt/raid/users/petr/workspace/ltr_finder_test/test_data/DANTE_filtered_part.gff3")
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
78 s <- readDNAStringSet("/mnt/raid/users/petr/workspace/ltr_finder_test/test_data/Rbp_part.fa")
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
79
9
1aa578e6c8b3 "planemo upload commit 9488b982bae902f1868785ec4ad47134dac50ff3"
petr-novak
parents: 8
diff changeset
80 # oriza
1aa578e6c8b3 "planemo upload commit 9488b982bae902f1868785ec4ad47134dac50ff3"
petr-novak
parents: 8
diff changeset
81 g <- rtracklayer::import("/mnt/raid/users/petr/workspace/dante_ltr/test_data/big_test_data/DANTE_full_oryza.gff3")
1aa578e6c8b3 "planemo upload commit 9488b982bae902f1868785ec4ad47134dac50ff3"
petr-novak
parents: 8
diff changeset
82 s <- readDNAStringSet("/mnt/raid/users/petr/workspace/dante_ltr/test_data/big_test_data/o_sativa_msu7.0.fasta")
1aa578e6c8b3 "planemo upload commit 9488b982bae902f1868785ec4ad47134dac50ff3"
petr-novak
parents: 8
diff changeset
83
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
84 g <- rtracklayer::import("/mnt/raid/users/petr/workspace/dante_ltr/test_data
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
85 /DANTE_Vfaba_chr5.gff3")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
86 s <- readDNAStringSet("/mnt/ceph/454_data/Vicia_faba_assembly/assembly/ver_210910
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
87 /fasta_parts/211010_Vfaba_chr5.fasta")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
88
2
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
89 g <- rtracklayer::import("/mnt/raid/users/petr/workspace/dante_ltr/test_data/big_test_data//Cocoa_theobroma_DANTE_filtered.gff3")
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
90 s <- readDNAStringSet("/mnt/raid/users/petr/workspace/dante_ltr/test_data/big_test_data/Cocoa_theobroma_chr1.fasta.gz")
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
91
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
92 source("R/ltr_utils.R")
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
93 ## feature distance model
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
94 FDM <- readRDS("./databases/feature_distances_model.RDS")
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
95 g <- rtracklayer::import("./test_data/sample_DANTE.gff3")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
96 s <- readDNAStringSet("./test_data/sample_genome.fasta")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
97 outfile <- "/mnt/raid/users/petr/workspace/ltr_finder_test/te_with_domains_2.gff3"
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
98 lineage_info <- read.table("databases/lineage_domain_order.csv", sep = "\t", header =
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
99 TRUE, as.is = TRUE)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
100 trna_db <- "./databases/tRNAscan-SE_ALL_spliced-no_plus-old-tRNAs_UC_unique-3ends.fasta"
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
101
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
102 }
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
103
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
104 # MAIN #############################################################
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
105
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
106 # load data:
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
107
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
108 cat("reading gff...")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
109 g <- rtracklayer::import(opt$gff3, format = 'gff3') # DANTE gff3
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
110 cat("done\n")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
111 cat("reading fasta...")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
112 s <- readDNAStringSet(opt$reference_sequence) # genome assembly
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
113 cat("done\n")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
114 outfile <- opt$output
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
115 # clean sequence names:
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
116 names(s) <- gsub(" .+", "", names(s))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
117 lineage_domain <- lineage_info$Domains.order
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
118 lineage_domain_span <- lineage_info$domain_span
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
119 lineage_ltr_mean_length <- lineage_info$ltr_length
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
120 lineage_offset5prime <- lineage_info$offset5prime
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
121 lineage_offset3prime <- lineage_info$offset3prime
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
122 ln <- gsub("ss/I", "ss_I", gsub("_", "/", gsub("/", "|", lineage_info$Lineage)))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
123 names(lineage_offset3prime) <- ln
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
124 names(lineage_offset5prime) <- ln
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
125 names(lineage_domain) <- ln
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
126 names(lineage_domain_span) <- ln
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
127 names(lineage_ltr_mean_length) <- ln
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
128 lineage_domains_sequence <- unlist(mapply(function(d,l) {
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
129 paste(strsplit(d, " ")[[1]], ":", l, sep = "")
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
130 }, d = lineage_domain, l = names(lineage_domain)))
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
131
8
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
132 # filter g gff3
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
133 g <- dante_filtering(g, Relative_Length = opt$min_relative_length) # default
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
134
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
135 seqlengths(g) <- seqlengths(s)[names(seqlengths(g))]
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
136 g <- add_coordinates_of_closest_neighbor(g)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
137
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
138 # add info about domain order:
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
139 g$domain_order <- as.numeric(factor(paste(g$Name, g$Final_Classification, sep = ":"), levels = lineage_domains_sequence))
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
140 # set NA to 0 in g$domain_order ( some domains are not fromm ClassI elements
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
141 g$domain_order[is.na(g$domain_order)] <- 0
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
142
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
143 # NOTE - some operation is faster of GrangesList but some on list of data.frames
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
144 # this is primary clusteing
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
145 cls <- get_domain_clusters(g)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
146 gcl <- split(as.data.frame(g), cls)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
147 # gcl_as_GRL <- split(g, cls) # delete?
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
148
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
149 cls_alt <- get_domain_clusters_alt(g, FDM)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
150 g$Cluster <- as.numeric(factor(cls_alt))
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
151
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
152 gcl_alt <- split(as.data.frame(g), cls_alt)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
153
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
154 TE_partial <- GRanges(seqnames = sapply(gcl_alt, function(x) x$seqnames[1]),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
155 Name = sapply(gcl_alt, function(x) x$Final_Classification[1]),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
156 Final_Classification = sapply(gcl_alt, function(x) x$Final_Classification[1]),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
157 ID = sapply(gcl_alt, function(x) paste0("TE_partial_", x$Cluster[1])),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
158 strand = sapply(gcl_alt, function(x) x$strand[1]),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
159 Ndomains = sapply(gcl_alt, function(x) nrow(x)),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
160 type = "transposable_element",
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
161 source = "dante_ltr",
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
162 Rank="D",
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
163 IRanges(start = sapply(gcl_alt, function(x) min(x$start)),
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
164 end = sapply(gcl_alt, function(x) max(x$end)))
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
165 )
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
166 g$Ndomains_in_cluster <- count_occurences_for_each_element(g$Cluster)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
167 g$Parent <- paste0("TE_partial_", g$Cluster)
8
9de392f2fc02 "planemo upload commit d6433b48c9bae079edb06364147f19500501c986"
petr-novak
parents: 7
diff changeset
168 g$Rank <- "D"
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
169
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
170 # keep only partial TE with more than one domain
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
171 TE_partial_with_more_than_one_domain <- TE_partial[TE_partial$Ndomains > 1]
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
172 g_with_more_than_one_domain <- g[as.vector(g$Ndomains_in_cluster > 1)]
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
173
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
174 # first filtering - remove TEs with low number of domains
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
175 gcl_clean <- clean_domain_clusters(gcl, lineage_domain_span, min_domains = 5 - opt$max_missing_domains)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
176
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
177 # glc annotation
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
178 gcl_clean_lineage <- sapply(gcl_clean, function(x) x$Final_Classification[1])
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
179 gcl_clean_domains <- sapply(gcl_clean, function(x) ifelse(x$strand[1] == "-",
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
180 paste(rev(x$Name), collapse = " "),
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
181 paste(x$Name, collapse = " "))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
182 )
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
183
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
184 # compare detected domains with domains in lineages from REXdb database
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
185 dd <- mapply(domain_distance,
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
186 d_query = gcl_clean_domains,
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
187 d_reference = lineage_domain[gcl_clean_lineage])
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
188
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
189 # get lineages which has correct number and order of domains
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
190 # gcl_clean2 <- gcl_clean[gcl_clean_domains == lineage_domain[gcl_clean_lineage]]
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
191 gcl_clean2 <- gcl_clean[dd <= opt$max_missing_domains]
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
192
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
193 gcl_clean_with_domains <- gcl_clean2[check_ranges(gcl_clean2, s)]
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
194 gr <- get_ranges(gcl_clean_with_domains)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
195
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
196
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
197 cat('Number of analyzed regions:\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
198 cat('Total number of domain clusters : ', length(gcl), '\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
199 cat('Number of clean clusters : ', length(gcl_clean), '\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
200 cat('Number of clusters with complete domain set : ', length(gcl_clean_with_domains), '\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
201
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
202
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
203 te_strand <- sapply(gcl_clean_with_domains, function(x)x$strand[1])
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
204 te_lineage <- sapply(gcl_clean_with_domains, function(x)x$Final_Classification[1])
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
205
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
206 max_left_offset <- ifelse(te_strand == "+", lineage_offset5prime[te_lineage], lineage_offset3prime[te_lineage])
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
207 max_right_offset <- ifelse(te_strand == "-", lineage_offset5prime[te_lineage], lineage_offset3prime[te_lineage])
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
208
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
209 grL <- get_ranges_left(gcl_clean_with_domains, max_left_offset)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
210 grR <- get_ranges_right(gcl_clean_with_domains, max_right_offset)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
211
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
212 s_left <- getSeq(s, grL)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
213 s_right <- getSeq(s, grR)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
214
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
215 expected_ltr_length <- lineage_ltr_mean_length[sapply(gcl_clean_with_domains, function (x)x$Final_Classification[1])]
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
216 # for statistics
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
217 RT <- g[g$Name == "RT" & substring(g$Final_Classification, 1, 11) == "Class_I|LTR"]
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
218 # cleanup
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
219 #rm(g)
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
220 rm(gcl)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
221 rm(gcl_clean)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
222 rm(gcl_clean2)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
223
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
224 names(te_strand) <- paste(seqnames(gr), start(gr), end(gr), sep = "_")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
225 names(s_left) <- paste(seqnames(grL), start(grL), end(grL), sep = "_")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
226 names(s_right) <- paste(seqnames(grR), start(grR), end(grR), sep = "_")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
227 cat('Identification of LTRs...')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
228 TE <- mclapply(seq_along(gr), function(x)get_TE(s_left[x],
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
229 s_right[x],
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
230 gcl_clean_with_domains[[x]],
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
231 gr[x],
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
232 grL[x],
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
233 grR[x],
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
234 expected_ltr_length[x]),
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
235 mc.set.seed = TRUE, mc.cores = opt$cpu, mc.preschedule = FALSE
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
236 )
2
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
237
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
238 cat('done.\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
239
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
240 good_TE <- TE[!sapply(TE, is.null)]
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
241 cat('Number of putative TE with identified LTR :', length(good_TE), '\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
242
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
243 # TODO - extent TE region to cover also TSD
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
244 ID <- paste0('TE_', sprintf("%08d", seq(good_TE)))
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
245 gff3_list <- mcmapply(get_te_gff3, g = good_TE, ID = ID, mc.cores = opt$cpu)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
246
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
247 cat('Identification of PBS ...')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
248 gff3_list2 <- mclapply(gff3_list, FUN = add_pbs, s = s, trna_db = trna_db, mc.set.seed = TRUE, mc.cores = opt$cpu, mc.preschedule = FALSE)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
249 cat('done\n')
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
250 gff3_out <- do.call(c, gff3_list2)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
251
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
252 # define new source
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
253 src <- as.character(gff3_out$source)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
254 src[is.na(src)] <- "dante_ltr"
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
255 gff3_out$source <- src
2
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
256 gff3_out$Rank <- get_te_rank(gff3_out)
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
257
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
258 # add partial TEs but first remove all ovelaping
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
259 TE_partial_parent_part <- TE_partial_with_more_than_one_domain[TE_partial_with_more_than_one_domain %outside% gff3_out]
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
260 TE_partial_domain_part <- g[g$Parent %in% TE_partial_parent_part$ID]
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
261
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
262 gff3_out <- sort(c(gff3_out, TE_partial_domain_part, TE_partial_parent_part), by = ~ seqnames * start)
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
263
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
264 export(gff3_out, con = paste0(outfile, ".gff3"), format = 'gff3')
2
f131886ea194 "planemo upload commit 891bfe9acf7349c2b887aff6d7e52a7f4ebf3b3a"
petr-novak
parents: 0
diff changeset
265
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
266 all_tbl <- get_te_statistics(gff3_out, RT)
7
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
267 all_tbl <- cbind(Classification = rownames(all_tbl), all_tbl)
c33d6583e548 "planemo upload commit 50884f7f0269a0bbde078f24fe5020975693bcd9"
petr-novak
parents: 2
diff changeset
268 write.table(all_tbl, file = paste0(outfile, "_statistics.csv"), sep = "\t", quote = FALSE, row.names = FALSE)
0
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
269 # export fasta files:
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
270 s_te <- get_te_sequences(gff3_out, s)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
271 for (i in seq_along(s_te)) {
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
272 outname <- paste0(outfile, "_", names(s_te)[i], ".fasta")
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
273 writeXStringSet(s_te[[i]], filepath = outname)
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
274 }
7b0bbe7477c4 "planemo upload commit 92c684dff3b377c8c08654c7f3d46a133385e3e0-dirty"
petr-novak
parents:
diff changeset
275