annotate clean_rm_output.R @ 0:cf3cea0a3039 draft

Uploaded
author petr-novak
date Thu, 07 Oct 2021 06:07:34 +0000 (2021-10-07)
parents
children e955b40ad3a4
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
1 #!/usr/bin/env Rscript
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
2 suppressPackageStartupMessages(library(rtracklayer))
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
3
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
4 gff_cleanup = function(gff){
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
5 ## remove overlapin annotation track - assign new annot
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
6 gff_disjoin = disjoin(gff, with.revmap=TRUE)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
7 ## append annotation:
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
8 gff_names = mclapply(as.list(gff_disjoin$revmap), FUN = function(x)gff$Name[x], mc.cores = 8)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
9 gff_strands = mclapply(as.list(gff_disjoin$revmap), FUN = function(x)strand(gff[x]), mc.cores = 8)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
10 new_annot = sapply(sapply(gff_names, unique), paste, collapse="|")
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
11 strand_attribute = sapply(sapply(gff_strands, unique), paste, collapse="|")
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
12 gff_disjoin$strands=strand_attribute
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
13 gff_disjoin$source="RM"
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
14 gff_disjoin$type="repeat"
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
15 gff_disjoin$score=NA
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
16 gff_disjoin$phase=NA
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
17 gff_disjoin$Name=new_annot
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
18 gff_disjoin$revmap=NULL
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
19 return(gff_disjoin)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
20 }
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
21 infile = commandArgs(T)[1]
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
22 outfile = commandArgs(T)[2]
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
23
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
24 ## infile = "./test_data/raw_rm.out"
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
25
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
26 rm_out = read.table(infile, as.is=TRUE, sep="", skip = 2, fill=TRUE, header=FALSE)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
27
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
28 gff = GRanges(seqnames = rm_out$V5, ranges = IRanges(start = rm_out$V6, end=rm_out$V7))
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
29
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
30 # repeat class after # symbol - syntax 1
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
31 gff$Name=rm_out$V11
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
32
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
33 ## is repeat type is specifies by double underscore:
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
34 ## then rm_out$V11 is unspecified
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
35 if (any(rm_out$V11 == "Unspecified")){
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
36 ## set Name from prefix
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
37 ## TODO
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
38 inc = rm_out$V11 == "Unspecified"
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
39 Name = gsub("__.+","",rm_out$V10)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
40 # chanche Usnpsecified to new name
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
41 gff$Name[inc] = Name
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
42 }
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
43
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
44
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
45 ## join neighbors with the same annotation, disregard strand!
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
46 result <- unlist(reduce(split(gff, gff$Name)))
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
47 result$Name <- names(result)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
48
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
49 ## TODO
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
50 ## identify conflicting annotation, replace by LCA but keep origin list of classifications
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
51
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
52 gff_out = sortSeqlevels(result)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
53 gff_out = sort(gff_out)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
54 gff_out$type = "repeat_region"
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
55 gff_out$source = "RepeatMasker_parsed"
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
56 gff_out$ID=paste0(gff_out$Name, "_", seq_along(gff_out$Name))
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
57 export(gff_out, format = "gff3", con=outfile)
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
58
cf3cea0a3039 Uploaded
petr-novak
parents:
diff changeset
59