Skip to content

Latest commit

 

History

History
41 lines (31 loc) · 1.91 KB

File metadata and controls

41 lines (31 loc) · 1.91 KB

Context-specific modeling (tINIT / ftINIT)

Extract a tissue- or condition-specific model from a reference GEM plus gene scores derived from omics data. Two algorithms are provided in {mod}raven_python.init.

Scoring

Gene scores drive both algorithms. Build them from expression with {func}raven_python.init.gene_scores_from_expression and turn them into reaction scores via {func}raven_python.init.score_reactions_from_genes (a GPR walk shared with the omics adapters — see the omics guide).

tINIT

  • {func}raven_python.init.run_init — the classic INIT MILP (rewritten in optlang).
  • {func}raven_python.init.get_init_model — the full tINIT pipeline (dead-end removal → run_init).

ftINIT (faster, staged)

  • {func}raven_python.init.run_ftinit — the single-step ftINIT MILP (continuous indicators for positive-score reactions; binaries only on negatives — the speedup over run_init).
  • {func}raven_python.init.ftinit — the full pipeline: {func}raven_python.init.prep_init_model → staged run_ftinit → {func}raven_python.init.fill_tasks → {func}raven_python.init.remove_low_score_genes.

Tasks and defaults

ftINIT's task layer keeps essential metabolic tasks feasible; define and check tasks with the tasks guide. The parameter defaults (mip_gap, big_m, force_on, eps, prod_weight, scaling) and their robustness to noisy input are calibrated in the parameter-calibration study, and the equivalence to MATLAB RAVEN is established in the Human-GEM validation.

:::{important} Genome-scale (f)tINIT MILPs currently require Gurobi for tractable solve times; toy and unit-test problems run on GLPK. See the solver benchmark. Metabolomics-based scoring is the one piece not yet implemented (raises NotImplementedError). :::