Add reparametrize_log and petab1to2(reparametrize=...)

[dweindl]

PEtab v2 has no parameterScale. Converting a v1 parameterScaleNormal(μ, σ) prior on a log10-scaled parameter to log-normal(μ ln10, σ ln10) keeps the prior distribution, but adds a Jacobian term to the v2 MAP objective, which moves the optimum by −σ² ln10 decades. No v2 prior over the unscaled parameter can preserve both the distribution and the optimum.

This adds petab.v2.reparametrize_log, which estimates log10_<id> (or log_<id>) instead of <id> and defines <id> = 10^log10_<id>, as recommended by the v2 spec for log-scale priors:

• parameter table: log-transformed bounds and nominal value; priors from the priors argument, or derived from log-normal, log-laplace and log-uniform priors (other x-space priors raise)
• SBML model: new parameter log10_<id> and initial assignment <id> = 10^log10_<id> (replacing any existing one; <id> stays constant=true)
• observable/noise formulas and condition target values: <id> → 10^log10_<id>
• measurement overrides: replaced by the new IDs, numeric overrides are log-transformed, and the placeholder ph is replaced by 10^ph in the formulas

petab1to2 and petab_files_1to2 get a reparametrize argument:

• "none" (default): current behaviour, now with a warning for converted parameterScale* priors
• "priors": log-scaled parameters with parameterScale* priors
• "all": all estimated log-scaled parameters without x-space priors; this also preserves the v1 default prior (uniform on parameter scale)

Log-scaled parameters sharing a placeholder with selected parameters are added. With "all", parameters without parameter-scale priors are skipped (with a warning) if their placeholder can't be log-transformed (e.g., Bachmann has 0 offsets). Parameters with lowerBound = 0 (allowed in v1 for log-scaled parameters with an initialization prior) are skipped with a warning, as their lower bound on log scale would be -inf.

Checked on the benchmark collection: all 35 problems convert and pass v2 linting with "all" (now also tested in CI); with "priors", 44 parameters are reparametrized in Bachmann, Isensee, Raimundez and Schwen. Objective equivalence with a simulator is not checked here (to be done in AMICI / the benchmark repo).

Open point: default stays "none" until Benchmarking-Initiative/Benchmark-Models-PEtab#333 is decided.

References Benchmarking-Initiative/Benchmark-Models-PEtab#333

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

❌ Patch coverage is 96.25468% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.85%. Comparing base (5fb4066) to head (4be4d4a).

Files with missing lines	Patch %	Lines
petab/v2/reparametrize.py	95.37%	4 Missing and 4 partials ⚠️
petab/v2/petab1to2.py	97.84%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #539      +/-   ##
==========================================
+ Coverage   77.98%   78.85%   +0.87%     
==========================================
  Files          68       69       +1     
  Lines        7791     8055     +264     
  Branches     1390     1461      +71     
==========================================
+ Hits         6076     6352     +276     
+ Misses       1227     1213      -14     
- Partials      488      490       +2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.