cmatkhan commited on
Commit
c033470
·
verified ·
1 Parent(s): 650ca90

Upload pull_gse65682.R with huggingface_hub

Browse files
Files changed (1) hide show
  1. pull_gse65682.R +136 -0
pull_gse65682.R ADDED
@@ -0,0 +1,136 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Version info: R 4.2.2, Biobase 2.58.0, GEOquery 2.66.0, limma 3.54.0
2
+ ################################################################
3
+ # Data plots for selected GEO samples
4
+ library(GEOquery)
5
+ library(limma)
6
+ library(hgu219.db)
7
+ library(umap)
8
+ library(here)
9
+ library(tidyverse)
10
+
11
+ # load series and platform data from GEO
12
+
13
+ dir.create("data/gse65682")
14
+ gset <- getGEO("GSE65682", destdir = "data/gse65682", GSEMatrix = TRUE, getGPL = FALSE)
15
+ if (length(gset) > 1) idx <- grep("GPL13667", attr(gset, "names")) else idx <- 1
16
+ gset <- gset[[idx]]
17
+
18
+ ex <- exprs(gset)
19
+ # log2 transform
20
+ qx <- as.numeric(quantile(ex, c(0., 0.25, 0.5, 0.75, 0.99, 1.0), na.rm = T))
21
+ LogC <- (qx[5] > 100) ||
22
+ (qx[6] - qx[1] > 50 && qx[2] > 0)
23
+ if (LogC) {
24
+ ex[which(ex <= 0)] <- NaN
25
+ ex <- log2(ex)
26
+ }
27
+
28
+ # box-and-whisker plot
29
+ dev.new(width = 3 + ncol(gset) / 6, height = 5)
30
+ par(mar = c(7, 4, 2, 1))
31
+ title <- paste("GSE65682", "/", annotation(gset), sep = "")
32
+ boxplot(ex, boxwex = 0.7, notch = T, main = title, outline = FALSE, las = 2)
33
+ dev.off()
34
+
35
+ # expression value distribution plot
36
+ par(mar = c(4, 4, 2, 1))
37
+ title <- paste("GSE65682", "/", annotation(gset), " value distribution", sep = "")
38
+ plotDensities(ex, main = title, legend = F)
39
+
40
+ # mean-variance trend
41
+ ex <- na.omit(ex) # eliminate rows with NAs
42
+ plotSA(lmFit(ex), main = "Mean variance trend, GSE65682")
43
+
44
+ # UMAP plot (multi-dimensional scaling)
45
+ ex <- ex[!duplicated(ex), ] # remove duplicates
46
+ ump <- umap(t(ex), n_neighbors = 15, random_state = 123)
47
+ plot(ump$layout, main = "UMAP plot, nbrs=15", xlab = "", ylab = "", pch = 20, cex = 1.5)
48
+
49
+ clean_char <- function(x, prefix_regex) {
50
+ val <- str_squish(str_remove(x, prefix_regex))
51
+ na_if(val, "NA")
52
+ }
53
+
54
+ pdat_mars <- as_tibble(pData(gset), rownames = "sample_id") |>
55
+ select(sample_id, starts_with("characteristics"))
56
+
57
+ purrr::imap(pdat_mars |> select(-sample_id), ~ count(tibble(value = .x), value, name = "n") |> arrange(desc(n)))
58
+
59
+ gse65682_metadata <- as_tibble(pData(gset), rownames = "sample_id") |>
60
+ mutate(
61
+ healthy_control = str_detect(title, "healthy"),
62
+ title_date = str_extract(title, "\\d\\d_\\d{2}_\\d{4}"),
63
+ title_identifier1 = str_extract(title, "(?<=\\d{2}_\\d{2}_\\d{4}_)\\w\\d+"),
64
+ title_identifier2 = str_extract(title, "\\d+(?=\\])"),
65
+ sex = clean_char(characteristics_ch1, "^gender:\\s*"),
66
+ age = as.numeric(clean_char(characteristics_ch1.1, "^age:\\s*")),
67
+ pneumonia_diagnosis = clean_char(characteristics_ch1.2, "^pneumonia diagnoses:\\s*"),
68
+ thrombocytopenia = clean_char(characteristics_ch1.3, "^thrombocytopenia:\\s*"),
69
+ endotype_cohort = clean_char(characteristics_ch1.4, "^endotype_cohort:\\s*"),
70
+ endotype_class = clean_char(characteristics_ch1.5, "^endotype_class:\\s*"),
71
+ mortality_28d = as.logical(as.integer(clean_char(characteristics_ch1.6, "^mortality_event_28days:\\s*"))),
72
+ time_to_event_28d = as.numeric(clean_char(characteristics_ch1.7, "^time_to_event_28days:\\s*")),
73
+ icu_acquired_infection = clean_char(characteristics_ch1.8, "^icu_acquired_infection:\\s*"),
74
+ icu_acquired_infection_paired = clean_char(characteristics_ch1.9, "^icu_acquired_infection_paired:\\s*"),
75
+ diabetes_mellitus = clean_char(characteristics_ch1.10, "^diabetes_mellitus:\\s*"),
76
+ abdominal_sepsis_or_control = clean_char(characteristics_ch1.11, "^abdominal_sepsis_and_controls:\\s*")
77
+ ) |>
78
+ dplyr::select(
79
+ sample_id, title_identifier1, title_identifier2, healthy_control, sex,
80
+ age, pneumonia_diagnosis, thrombocytopenia, endotype_cohort, endotype_class,
81
+ mortality_28d, time_to_event_28d, icu_acquired_infection, icu_acquired_infection_paired,
82
+ diabetes_mellitus, abdominal_sepsis_or_control
83
+ )
84
+
85
+ stopifnot(gse65682_metadata |> filter(!is.na(endotype_class)) |> count(is.na(mortality_28d)) |> pull(n) == 479)
86
+
87
+ your_probes_mars <- rownames(exprs(gset))
88
+ biocu219_probes <- keys(hgu219.db, keytype = "PROBEID")
89
+
90
+ length(your_probes_mars)
91
+ length(biocu219_probes)
92
+ length(intersect(your_probes_mars, biocu219_probes))
93
+ stopifnot(mean(!your_probes_mars %in% biocu219_probes) == 0)
94
+
95
+ anno_u219 <- AnnotationDbi::select(
96
+ hgu219.db,
97
+ keys = your_probes_mars,
98
+ columns = c("SYMBOL", "ENTREZID", "ENSEMBL", "GENENAME", "UNIPROT"),
99
+ keytype = "PROBEID"
100
+ )
101
+
102
+ gse65682_feature_metadata <- anno_u219 |>
103
+ group_by(PROBEID) |>
104
+ reframe(
105
+ ENSEMBL = paste(unique(na.omit(ENSEMBL)), collapse = ";"),
106
+ ENTREZID = dplyr::first(ENTREZID),
107
+ SYMBOL = dplyr::first(SYMBOL),
108
+ GENENAME = dplyr::first(GENENAME),
109
+ UNIPROT = paste(unique(na.omit(UNIPROT)), collapse = ";")
110
+ ) |>
111
+ ungroup() |>
112
+ mutate(across(c(ENSEMBL, UNIPROT), ~ na_if(.x, ""))) |>
113
+ rename(feature_id = PROBEID) |>
114
+ right_join(tibble(feature_id = your_probes_mars), by = "PROBEID") # guarantee every probe gets a row
115
+
116
+ stopifnot(nrow(gse65682_feature_metadata) == length(your_probes_mars))
117
+ stopifnot(!any(duplicated(gse65682_feature_metadata$feature_id)))
118
+
119
+ mean(is.na(gse65682_feature_metadata$SYMBOL)) # for comparison against the other datasets' rates
120
+
121
+ gse65682_expr_long <- as_tibble(exprs(gset), rownames = "feature_id") |>
122
+ pivot_longer(-feature_id, names_to = "sample_id", values_to = "value") |>
123
+ arrange(sample_id, feature_id)
124
+
125
+ stopifnot(nrow(gse65682_expr_long) == nrow(exprs(gset)) * ncol(exprs(gset)))
126
+ stopifnot(all(unique(gse65682_expr_long$sample_id) %in% gse65682_metadata$sample_id))
127
+ stopifnot(all(unique(gse65682_expr_long$PROBEID) %in% gse65682_feature_metadata$PROBEID))
128
+
129
+ stopifnot(!any(duplicated(gse65682_metadata$sample_id)))
130
+ stopifnot(nrow(gse65682_metadata) == 802)
131
+
132
+ dir.create("data/gse65682/parquet", recursive = TRUE, showWarnings = FALSE)
133
+
134
+ arrow::write_parquet(gse65682_metadata, "data/gse65682/parquet/sample_metadata.parquet")
135
+ arrow::write_parquet(gse65682_feature_metadata, "data/gse65682/parquet/feature_metadata.parquet")
136
+ arrow::write_parquet(gse65682_expr_long, "data/gse65682/parquet/expression.parquet")