Guide#

Short, task-focused pages: one job per page, in MATLAB and Python side by side. Start at Introduction if the ecModel concept is new, or Getting started to jump straight into loading a model, and read on from there, or go directly to whichever task is at hand.

GECKO 4.0.0b1

This Guide documents GECKO 4, a pre-release still under development. Pages call out functions that are new in GECKO 4 and were not part of the published GECKO 3.0 Nature Protocols pipeline; see GECKO 3 → GECKO 4 for the full list of changes, and for how to stay on GECKO 3 if a workflow depends on its behavior.

MATLAB and Python

Every task page shows both languages side by side: GECKO (MATLAB), built on RAVEN, and geckopy (Python), built on cobrapy. Function names differ only mechanically between the two (camelCase vs. snake_case); see MATLAB ↔ Python for the full name mapping, and API reference for every function's signature.

Text outside a MATLAB/Python tab reads the same whichever language is selected. Where it must refer to the model's enzyme data, ec.kcat stands for ecModel.ec.kcat in MATLAB and ec_model.ec.kcat in Python; field names are identical in both languages except rxnEnzMat (Python: rxn_enz_mat).

Pages#

Start

Introduction to ecModels, what an ecModel is and how enzyme constraints are added to a conventional GEM.
Installation, set up GECKO or geckopy.

Guide, the steps that build and analyze an ecModel, in order:

Getting started, scaffold a project and load a first model.
Building an ecModel, expand a conventional GEM into ecModel structure.
Gathering kcat values, collect turnover numbers from BRENDA and DLKcat.
Applying kcat values, write collected kcat values into the ecModel's enzyme constraints.
Growth-rate tuning, close the gap between a freshly built ecModel's growth rate and the organism's actual one.
Tuning against experimental data, curate the kcat values tuning identified as limiting.
Proteomics integration, constrain individual enzyme concentrations from measured proteomics data.
Relaxing overconstrained proteomics, two ways to loosen proteomics constraints that leave the ecModel unable to reach its intended growth rate.
Simulation and analysis, objective choice, flux variability, sampling, mapping ecModel fluxes back to a conventional GEM, and context-specific ecModels.
Enzyme usage and bottlenecks, which enzymes a solution uses, and which ones are actually limiting the objective.

Background, reference material the Guide pages above link out to rather than repeat:

GECKO light vs. full ecModels, the two ecModel layouts, what light gives up, and when to choose each.
ecModel YAML format, the file format both toolboxes read and write, field by field.
Files in the data folder, every file a project's data folder can hold.
Model adapter parameters, every model adapter parameter, in both languages.

Approximate timing#

Rough wall-clock time for a first pass through each Guide step, on a genome-scale model such as yeast-GEM. Actual timing depends heavily on model size, internet connection, and which solver is configured.

Pages	Approx. timing	Notes
Getting started, Building an ecModel	15 min
Gathering kcats, Applying kcats	1 h	DLKcat prediction accounts for most of it
Growth-rate tuning, Tuning against experimental data	15 min
Proteomics integration	15 min
Relaxing overconstrained proteomics	15 min	can take longer with lower-quality proteomics data, more enzymes need flexibilizing
Simulation and analysis	3 h	ecFVA at genome scale accounts for most of it, and is far shorter with a commercial solver configured
Enzyme usage and bottlenecks	15 min

Guide#

Pages#

Approximate timing#

See also#