Data on the runner

A cohort of whole genomes does not go through the browser. Put it in the daemon's workspace, under data/, and the studio lists it as records while the runner is connected.

~/sigscape/
  data/
    pcawg/
      snv/            2,703 VCFs → one record, "snv", format VCF folder
      exposures.csv   → one record, "exposures", an exposure table
  uploads/            what the page sent
  sigscape-runs/      every run, one folder each

• A folder that holds VCFs is one record of format VCF folder; the tool reads the folder on the machine.
• A table small enough to read (under 8 MB) is fetched and parsed like a drop, so an exposure table or an Add object JSON is the same record it would be from your disk, with its workspace path attached.
• Larger files are listed unparsed and passed by path.

Runner records appear in the data panel under the folder's name with a runner tag. Select them like any record. A tool that gets one passes its path to the daemon instead of uploading.

SigMA2 from a cohort

Select a VCF folder and an exposure table and open SigMA2: the Curate stage gets a Build object step (tumor type, signatures of interest, panels, reference genome, clustering), Simulate gets the real panel downsampling on the chosen build, and Tune trains on the chosen simulation. Curation made in the editor is applied before a simulation runs. Reference genomes come from sigma2's data directory; sigma2 data genome hg19 fetches one.