-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathconfig.R
162 lines (154 loc) · 4.09 KB
/
config.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
### options
shinyOptions(cache = diskCache("./app-cache", max_size = 100 * 1024^2))
options(readr.num_columns = 0)
options(stringsAsFactors = FALSE)
options(spinner.type = 6)
theme_set(theme_cowplot())
stack_size <- getOption("pandoc.stack.size", default = "1000m")
options(repos = BiocManager::repositories()) # for pushing to shinyapps.io
# options(shiny.reactlog = TRUE) # for checking shiny logic
### folders
rpath <- "R" # additional R code
datapath <- "data" # data, previewed and for download
annotpath <- "annot" # annotations from external sources
listpath <- "data/lists" # csv and txt lists used by Venn and Circos
### general data settings
apptitle_short <- "squirrelBox"
apptitle <- "13-lined ground squirrel hibernating liver and brain sequencing"
url <- "https://github.com/rnabioco/squirrelbox/"
s3 <- "https://squirrelbox.s3-us-west-2.amazonaws.com/zip6/squirrelbox6.tar.gz"
docker <- "https://hub.docker.com/r/raysinensis/squirrelbox"
versionN <- "1.2.0"
geoN <- "GSE106947"
genomeN <- "NCBI Assembly HiC_Itri_2"
genomeL <- "https://www.ncbi.nlm.nih.gov/assembly/GCA_016881025.1"
manuscriptN <- "Dynamic RNA Regulation in the Brain Underlies Physiological Plasticity in a Hibernating Mammal"
manuscriptN2 <- "Liver transcriptome dynamics during hibernation are shaped by a shifting balance between transcription and RNA stability"
manuscriptL <- "https://www.frontiersin.org/articles/10.3389/fphys.2020.624677"
manuscriptL2 <- "https://www.frontiersin.org/articles/10.3389/fphys.2021.662132"
annotN <- "new annotation"
annotL <- "ftp://ftp.ncbi.nlm.nih.gov/genomes/all/GCA/016/881/025/GCA_016881025.1_HiC_Itri_2/GCA_016881025.1_HiC_Itri_2_genomic.gtf.gz"
bsgenomeN <- "BSgenome.Itridecemlineatus.HiC_Itri_2"
bsgenomeL <- "https://squirrelbox.s3-us-west-2.amazonaws.com/BSgenome/BSgenome.SQ1_1.0.tar.gz"
pageN <- 10 # number of lines, for tables
warningN <- 100 # number of genes, for throwing warnings in line and heat plots
plot_width <- 8
plot_height <- 6
set_shinytheme <- "paper"
track_name <- "hub_1512849_KG_HiC"
track_url <- "https://squirrelhub.s3-us-west-1.amazonaws.com/hub/hub.txt"
gmt_file <- "c5.all.v7.1.symbols.gmt"
gmt_short <- "GO_"
sig_cut <- 0.001
ncore <- parallel::detectCores() - 1
start_tutorial <- TRUE # whether to start loaded app with tutorial
start_tabhint <- TRUE # whether to pop up message on first switch to tab
verbose_bench <- FALSE
chrlimit <- 17
qc_report <- "multiqc_report.html"
### choose and order columns
table_cols <- c(
"gene_id", # comment out to remove from table outputs
"unique_gene_symbol",
"clean_gene_symbol",
"original_gene_name",
"source"
)
orf_cols_join <- c(
"gene_id",
"chrom",
"source",
"orf_len",
"micropeptide_pred",
"micropeptide_homology",
"exons",
"rna_len",
"transcript_id",
"novel",
"min_padj",
"domains",
"br_expr",
"edited",
"majiq_directed"
)
orf_cols <- c(
"gene_id",
"transcript_id",
"rna_len",
"orf_len",
"micropeptide_pred",
"exons",
"edited"
)
maj_cols <- c(
"region",
"comp_pair",
"significant_90",
"significant_99",
"LSV_ID",
"A5SS",
"A3SS",
"ES"
)
### sample settings, define state colors and order, region order
state_cols <- c(
SA = rgb(255, 0, 0, maxColorValue = 255),
IBA = rgb(67, 205, 128, maxColorValue = 255),
Ent = rgb(155, 48, 255, maxColorValue = 255),
ET = rgb(120, 10, 130, maxColorValue = 255),
LT = rgb(25, 25, 112, maxColorValue = 255),
EAr = rgb(0, 0, 255, maxColorValue = 255),
Ar = rgb(0, 0, 255, maxColorValue = 255),
LAr = rgb(0, 102, 102, maxColorValue = 255),
SpD = rgb(255, 165, 0, maxColorValue = 255)
)
state_order <- c(
"SA",
"IBA",
"Ent",
"ET",
"LT",
"EAr",
"Ar",
"LAr",
"SpD"
)
region_main2 <- c(
"Forebrain",
"Hypothalamus",
"Medulla",
"Adrenal",
"Kidney"
)
region_main <- c(
"Liver",
"Liver_GRO-seq",
"Liver_GRO-seq_promoter"
)
region_order <- c(region_main, region_main2)
region_short <- c(
"liv",
"gro",
"gropro",
"fb",
"hy",
"med",
"adr",
"kid"
)
region_short_main <- c(
"liv",
"gro",
"gropro"
)
region_one <- c(
"l",
"g",
"p",
"f",
"h",
"m",
"a",
"k"
)
hide_region <- c("Adrenal", "Kidney")