Keep SRA records; remove only withdrawn Series¶
Amended by ADR 0008: the refresh keeps every SRA record, and a yearly maintenance run removes the withdrawn ones.
A refresh removes a GEO Series that GEO no longer lists, with its Series-keyed rows, after a fetch of it alone confirms it is gone. It never removes an SRA experiment or run, and it does not drop records no live Series reaches. liulab-data and its consumers reach SRA through Series, so an SRA record no live Series names is harmless, while confirming and pruning SRA records cost about an hour a refresh.
Consequences¶
- The store keeps SRA records NCBI no longer serves. A direct lookup of one returns its old
record instead of
NotFound. - A live Series whose Sample still names a suppressed experiment keeps listing it until GEO updates the Series.