Variability Configuration

See also

Related research paper: Journal of Pharmacological and Toxicological Methods

In order for variability data to appear in simulation results in the UIs of both the client and client-direct components, ApPredict needs to have made available to it variability lookup tables - a manifest of which is available from Gary Mirams’ repository.

Currently (as of Jan. 2018), whenever ApPredict is run with the --credible-intervals (which itself had its functionality extended May 2018 to allow the specification of credible interval percentile values) and --model <model id> args supplied, it will look in ApPredict’s current working directory [1] for an unpacked version [2] of the relevant lookup table. If it is not found ApPredict will attempt to download the packed file from the aforementioned repository and if such a file exists it will be downloaded, unpacked locally, and loaded into memory [3].

Footnotes

[1]

This has been achieved by modifying app-manager’s prepare.sh file to symlink to manually downloaded unpacked variability lookup tables in ApPredict’s current working directory. Ideally it is the *_BINARY.arch files that should be made available (as these are created dynamically from the non-BINARY .arch files when ApPredict loads them), although there may be compatibility issues using legacy versions derived from historical compilations.

[2]	Conventionally the variability lookup tables are released in .tgz format due to their large (e.g. 700Mb+) size.

[3]	This has the potential to drain the hardware’s available RAM which may impact server performance.