P2Mdb / code /read_data.R
kmmuleelab's picture
Reconcile submission source tables and document historical comparison limits
3d163e2 verified
Raw History Blame Contribute Delete
3.59 kB
# Rscript read_release.R PATH_TO_RDS [MATRIX_NAME] [PLATFORM]
# Read metadata and select confirmed specimen groups without rerunning analysis.
args <- commandArgs(trailingOnly = TRUE)
if (!length(args)) stop('Provide the path to a released RDS')
if (requireNamespace('SeuratObject', quietly = TRUE))
suppressPackageStartupMessages(library(SeuratObject))
x <- readRDS(args[1])
if (inherits(x, 'Seurat')) {
metadata <- x@meta.data
ids <- colnames(x)
stopifnot(identical(ids, rownames(metadata)), !anyDuplicated(ids))
cat('Single-cell object:', nrow(x), 'features;', length(ids), 'cells\n')
fields <- intersect(c('sample_id', 'patient_id', 'sample_class', 'specimen_role',
'sampled_tissue', 'celltype_article', 'celltype_singler',
'celltype_sctype'), names(metadata))
print(head(metadata[, fields, drop = FALSE]))
} else if (is.list(x) && all(c('expression', 'samples') %in% names(x))) {
expression <- x[['expression']]
metadata <- x[['samples']]
if (is.list(expression) && !is.data.frame(expression)) {
if (length(args) < 2L || !args[2] %in% names(expression)) {
print(names(expression))
stop('Multiple matrices: provide one exact MATRIX_NAME as argument 2')
}
expression <- expression[[args[2]]]
}
if (length(args) >= 3L) {
if (!'platform' %in% names(metadata)) stop('Platform metadata unavailable')
metadata <- metadata[!is.na(metadata$platform) & metadata$platform == args[3], , drop = FALSE]
}
ids <- colnames(expression)
if (anyDuplicated(metadata$sample_id))
stop('Sample IDs recur across platforms; provide the exact PLATFORM as argument 3')
idx <- match(ids, metadata$sample_id)
stopifnot(!anyNA(idx), !anyDuplicated(ids))
metadata <- metadata[idx, , drop = FALSE]
stopifnot(identical(ids, as.character(metadata$sample_id)))
cat('Bulk matrix:', nrow(expression), 'features;', length(ids), 'expression columns\n')
fields <- intersect(c('sample_id', 'platform', 'patient_id', 'sample_class',
'sampled_tissue', 'classification_source', 'sample_type_code',
'clinical_m_stage'), names(metadata))
print(head(metadata[, fields, drop = FALSE]))
print(x[['information']])
} else stop('Unsupported RDS structure')
if (!'sample_class' %in% names(metadata)) stop('Specimen class is not recorded')
print(table(metadata$sample_class, useNA = 'ifany'))
if (inherits(x, 'Seurat')) {
if (!'specimen_role' %in% names(metadata))
stop('Detailed specimen roles are unavailable; inspect samples.csv before selecting tissue groups')
primary_ids <- ids[which(metadata$specimen_role == 'primary_tumor')]
metastasis_ids <- ids[which(metadata$specimen_role == 'metastatic_tumor')]
} else {
primary_ids <- ids[which(metadata$sample_class == 'Primary')]
metastasis_ids <- ids[which(metadata$sample_class == 'Metastasis')]
}
cat('Tissue-group selections from recorded metadata:', length(primary_ids), 'Primary;',
length(metastasis_ids), 'Metastasis\n')
# Fluid specimens (including malignant ascites), normal and unresolved roles
# are excluded from single-cell tissue selections, even if sample_class is Metastasis.
# TCGA clinical M0/M1 is not substituted for specimen class.
# Available author, SingleR and scType annotations remain independent fields.
# Inspect expression scale, patient dependence and analysis/comparisons.csv
# before constructing any new contrast. This example performs no DEG analysis.