Assessment of missing cell-level metadata in single cell GEO records

Current GEO query used:

"expression profiling by high throughput sequencing" AND
("single nuclei" OR "single cell" OR "scRNAseq" OR "scRNA-seq" OR "snRNAseq" OR "snRNA-seq")

#> [1] "022122"
#> Response [https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=gds&query_key=1&WebEnv=MCID_6212fe0866927a572d71240b]
#>   Date: 2022-02-21 02:50
#>   Status: 200
#>   Content-Type: text/plain; charset=UTF-8
#>   Size: 6.52 MB
#> <ON DISK>  /home/runner/work/someta/someta/inst/extdata/022122/gds_result_022122.txt

Parse GEO query

Total entries: 7805

Number of entries filtered out because key words were not found: 926

Merged super and subseries: 593

The fraction of GEO entries with potential metadata (file with “meta”, “annot”, “clustering”, “colData”, or “type” in filename or rda/rds/rdata/h5ad/loom files) is 0.1760905. Note however that these terms include some false positives (such as gene annotation file, patient metadata, and phenotype table), which we manually inspected and corrected (false positive fraction at 0.0491803). Final fraction: 0.1726979

Comparison of parsed data to database from Svensson et al.

Overlap of scRNA-seq GEO entries between manual curation and the GEO query is 0.9658344. Number of entries not public: 0. Fraction with metadata: 0.2278665.

GEO cell-level metadata availability by year

GEO cell-level metadata availability by journal

Potential effects of GEO cell-level metadata on data reuse and citations

Investigation of whether metadata deposition is better in projects with authors developing scRNA-seq informatic tools (split out to “upload author/other author/not”, or just “author/not”)

Rui Fu