pull_dilgom.R

library(ArrayExpress)
library(illuminaHumanv3.db)
library(tidyverse)
library(here)

options(timeout = 1200)

outdir <- "data/dilgom/E-TABM-1036"
dir.create(outdir, recursive = TRUE, showWarnings = FALSE)

out_dilgom <- ArrayExpress::getAE("E-TABM-1036", path = outdir)

sdrf <- janitor::clean_names(read_tsv(out_dilgom$sdrf))

dilgom_metadata <- tibble(
    sample_id     = sdrf$source_name,
    accession     = "E-TABM-1036",
    age_band      = sdrf$characteristics_age,
    sex           = sdrf$characteristics_sex
)


expr_mat_colnames <- c("IlluminaID", str_split(readLines(out_dilgom$processedFiles, n = 1), "\t")[[1]][-1])
expr_mat_dilgom <- read_tsv(out_dilgom$processedFiles,
    skip = 2,
    col_names = expr_mat_colnames
)

dilgom_expr_long <- expr_mat_dilgom |>
    pivot_longer(-IlluminaID, names_to = "sample_id", values_to = "value") |>
    arrange(sample_id, IlluminaID)

# get probeset metadata

your_probes_v3 <- expr_mat_dilgom$IlluminaID

con_v3 <- illuminaHumanv3_dbconn()
extra_v3 <- DBI::dbGetQuery(con_v3, "SELECT * FROM ExtraInfo")

anno_v3 <- AnnotationDbi::select(
    illuminaHumanv3.db,
    keys = your_probes_v3,
    columns = c("SYMBOL", "ENTREZID", "ENSEMBL", "GENENAME", "UNIPROT"),
    keytype = "PROBEID"
)


anno_v3_collapsed <- anno_v3 |>
    group_by(PROBEID) |>
    summarise(
        ENSEMBL = paste(unique(na.omit(ENSEMBL)), collapse = ";"),
        ENTREZID = first(ENTREZID),
        SYMBOL = first(SYMBOL),
        GENENAME = first(GENENAME),
        UNIPROT = paste(unique(na.omit(UNIPROT)), collapse = ";"),
        .groups = "drop"
    ) |>
    mutate(across(c(ENSEMBL, UNIPROT), ~ na_if(.x, "")))

extra_v3_selected <- extra_v3 |>
    as_tibble() |>
    filter(IlluminaID %in% your_probes_v3) |>
    distinct(IlluminaID, .keep_all = TRUE) |>
    select(
        IlluminaID, ProbeQuality, CodingZone, ProbeSequence,
        SecondMatches, OtherGenomicMatches, RepeatMask,
        OverlappingSNP, GenomicLocation
    )

dilgom_feature_metadata <- tibble(IlluminaID = your_probes_v3) |>
    left_join(extra_v3_selected, by = "IlluminaID") |>
    left_join(anno_v3_collapsed, by = c("IlluminaID" = "PROBEID"))

# Check

stopifnot(nrow(dilgom_feature_metadata) == length(your_probes_v3))
stopifnot(!any(duplicated(dilgom_feature_metadata$IlluminaID)))

# Expected row count
stopifnot(nrow(dilgom_expr_long) == nrow(expr_mat_dilgom) * (ncol(expr_mat_dilgom) - 1))

# Every sample in expression data has a metadata row, and vice versa
expr_samples_dilgom <- dilgom_expr_long |> distinct(sample_id)
meta_samples_dilgom <- dilgom_metadata |> distinct(sample_id)

nrow(anti_join(expr_samples_dilgom, meta_samples_dilgom, by = "sample_id")) # expect 0
nrow(anti_join(meta_samples_dilgom, expr_samples_dilgom, by = "sample_id")) # expect 0 -- no documented QC exclusions here, unlike GAinS

# No duplicate sample rows in metadata
stopifnot(!any(duplicated(dilgom_metadata$sample_id)))

# feature_metadata is one row per unique probe, no fan-out
stopifnot(nrow(dilgom_feature_metadata) == length(your_probes_v3))
stopifnot(!any(duplicated(dilgom_feature_metadata$IlluminaID)))

# Every probe in expression data has an annotation row
stopifnot(all(unique(dilgom_expr_long$IlluminaID) %in% dilgom_feature_metadata$IlluminaID))

# No unexpected NAs in join keys
stopifnot(!anyNA(dilgom_expr_long$IlluminaID), !anyNA(dilgom_expr_long$sample_id))
stopifnot(!anyNA(dilgom_metadata$sample_id))

# write out

# dir.create("data/dilgom/parquet", recursive = TRUE, showWarnings = FALSE)

arrow::write_parquet(dilgom_metadata, "~/projects/hf_sepsis_collection/dilgom/sample_metadata.parquet")

arrow::write_parquet(dilgom_feature_metadata, "~/projects/hf_sepsis_collection/dilgom/feature_metadata.parquet")

arrow::write_parquet(dilgom_expr_long, "~/projects/hf_sepsis_collection/dilgom/expression.parquet")