Tellurium: a Calzone model archive and a paper-figure comparison

On this page

Execute the exact COMBINE archive cited in Figure4 and reference54 of Medley et al. (2018). It packages the Calzone2007 Drosophila model plus two SED-ML simulations. The author repository's former demos/ path moved to real-models/; the source manifest pins commiteff025846c5b12ee4350f043623332158e05f262 and the archive hash.

Repeat the run#

micromamba create -y -p ./env -f environment.yml
./env/bin/python fetch_sources.py
LD_LIBRARY_PATH="$PWD/env/lib" MPLBACKEND=Agg ./env/bin/python run_calzone.py --output ./run
LD_LIBRARY_PATH="$PWD/env/lib" ./env/bin/python compare_calzone.py --run ./run --output ./comparison

requirements-tested.txt and environment-explicit.txt record tested Python3.11/Tellurium2.2.13.1 dependencies. This is a current runtime rather than the paper-era environment. On this Linux conda runtime, libRoadRunner needed the environment's lib directory in LD_LIBRARY_PATH; Tellurium also required the undeclared six dependency. Those corrections affect the isolated runtime only.

The archive and its SBML/SED-ML are unchanged. executeCombineArchive runs the original and String-turnover variants with KiSAO0000019, each with1000 intervals(1001 samples), ending at250 and180 model-time units. The variant changes ksstg to0.02,kdstg to0.015 and initialStgPc to0, matching the article's described changes. All21 exported data-generators were finite. The SBML has a unit definition named min but no explicit model-level timeUnits attribute; outputs retain model-time units without an imposed conversion.

Comparison to Figure4B#

Before any archive execution,10 Wee1c points were selected from the published Figure4B image at times25,50,75,100,125,145,180,200,220 and240. Pixel calibration and the rule are in digitization-protocol.json. The bounds RMSE<=0.03 and maximum absolute error<=0.06 in model concentration units were chosen to allow figure line thickness/calibration and current deterministic-solver differences. These are OSRI comparison bounds, not tolerances published by the authors.

Target Observed Bound Result
Wee1c RMSE,10 selected points 0.00267333 0.03 Within bound
Wee1c maximum absolute error 0.00626005 0.06 Within bound

The comparison covers the selected Figure4B Wee1c trace. It does not establish agreement for all six panels, all parameters, the original BIOCHAM implementation, or biological prediction. Full SED-ML arrays and six generated plots remain retained. The compact CSV includes the entire1001-sample Figure4B trajectories; comparison-points.csv gives paper values, computed values and residuals. The paper's reference figure is open access under CC BY terms; the original archive/model retain their own source and attribution terms and are fetched rather than relicensed here.

The first export wrapper missed the nested dataGenerators dictionary; the archive itself executed. That partial export is retained, and the corrected export wrapper reruns the unchanged archive. See the run record and setup-attempt notes.

Status: Selected paper target compared within predeclared bounds using author archive/current runtime. Human scientific review of digitization, tolerance and scope is pending; no blanket reproduction of the full paper is claimed.

Reference-helper packaging check#

The original 10 reference points and tolerance protocol were generated by the working one-off digitize_references.py before any archive execution. The standalone digitize_reference.py was derived later for this portable bundle. Its first packaged copy lost a comment marker and failed to parse; that broken copy is retained as a packaging failure. Adding the marker fixed syntax only. The corrected helper ran in a fresh directory and regenerated the original 10-point CSV and protocol byte for byte. See packaging-validation.json for source and result hashes. No target, calibration, tolerance or model output changed.

Download the recorded files#

Complete run record · All candidate attempts. Human scientific review is pending.

View source on GitHub