Constraint-based metabolic modeling (COBRA). FBA, FVA, gene knockouts, flux sampling, SBML models, for systems biology and metabolic engineering analysis.
SKILL.md
COBRApy - Constraint-Based Reconstruction and Analysis
Overview
COBRApy is a Python library for constraint-based reconstruction and analysis (COBRA) of metabolic models, essential for systems biology research. Work with genome-scale metabolic models, perform computational simulations of cellular metabolism, conduct metabolic engineering analyses, and predict phenotypic behaviors.
Loading, building, or exporting genome-scale metabolic models (SBML, JSON, YAML)
Running FBA, pFBA, FVA, or flux sampling on COBRA models
Performing gene or reaction knockout screens and production envelope analysis
Designing or optimizing growth media and exchange constraints
Gap-filling infeasible models or validating model consistency
Installation
uv pip install "cobra==0.31.1"
MATLAB model I/O (optional):
uv pip install "cobra[array]==0.31.1"
COBRApy uses optlang for solvers. GLPK installs automatically via swiglpk. For large MILPs/QPs, cobra 0.29+ adds a hybrid solver (HIGHS/OSQP); model.solver = "osqp" now routes through hybrid and may error on plain LPs in a future release—prefer model.solver = "hybrid" when available.
Core Capabilities
COBRApy provides comprehensive tools organized into several key areas:
1. Model Management
Load existing models from repositories or files:
from cobra.io import load_model
# Bundled locally (no network): textbook, iJO1366, salmonella
model = load_model("textbook") # alias for e_coli_core (95 reactions)
model = load_model("e_coli_core") # same core E. coli model
model = load_model("iJO1366") # genome-scale E. coli (bundled)
model = load_model("salmonella") # Salmonella iYS1720 (bundled)
# Remote (BiGG / BioModels; requires network, cached after first fetch)
model = load_model("iML1515") # E. coli genome-scale on BiGG
# Load from files
from cobra.io import read_sbml_model, load_json_model, load_yaml_model
model = read_sbml_model("path/to/model.xml")
model = load_json_model("path/to/model.json")
model = load_yaml_model("path/to/model.yml")
Save models in various formats:
from cobra.io import write_sbml_model, save_json_model, save_yaml_model
write_sbml_model(model, "output.xml") # Preferred format
save_json_model(model, "output.json") # For Escher compatibility
save_yaml_model(model, "output.yml") # Human-readable
2. Model Structure and Components
Access and inspect model components:
# Access components
model.reactions # DictList of all reactions
model.metabolites # DictList of all metabolites
model.genes # DictList of all genes
# Get specific items by ID or index
reaction = model.reactions.get_by_id("PFK")
metabolite = model.metabolites[0]
# Inspect properties
print(reaction.reaction) # Stoichiometric equation
print(reaction.bounds) # Flux constraints
print(reaction.gene_reaction_rule) # GPR logic
print(metabolite.formula) # Chemical formula
print(metabolite.compartment) # Cellular location
from cobra.flux_analysis import pfba
solution = pfba(model)
Geometric FBA (find central solution):
from cobra.flux_analysis import geometric_fba
solution = geometric_fba(model)
4. Flux Variability Analysis (FVA)
Determine flux ranges for all reactions:
from cobra.flux_analysis import flux_variability_analysis
# Standard FVA
fva_result = flux_variability_analysis(model)
# FVA at 90% optimality
fva_result = flux_variability_analysis(model, fraction_of_optimum=0.9)
# Loopless FVA (eliminates thermodynamically infeasible loops)
fva_result = flux_variability_analysis(model, loopless=True)
# FVA for specific reactions
fva_result = flux_variability_analysis(
model,
reaction_list=["PFK", "FBA", "PGI"]
)
5. Gene and Reaction Deletion Studies
Perform knockout analyses:
from cobra.flux_analysis import (
single_gene_deletion,
single_reaction_deletion,
double_gene_deletion,
double_reaction_deletion
)
# Single deletions
gene_results = single_gene_deletion(model)
reaction_results = single_reaction_deletion(model)
# Double deletions (uses multiprocessing)
double_gene_results = double_gene_deletion(
model,
processes=4 # Number of CPU cores
)
# Manual knockout using context manager
with model:
model.genes.get_by_id("b0008").knock_out()
solution = model.optimize()
print(f"Growth after knockout: {solution.objective_value}")
# Model automatically reverts after context exit
6. Growth Media and Minimal Media
Manage growth medium:
# View current medium
print(model.medium)
# Modify medium (must reassign entire dict)
medium = model.medium
medium["EX_glc__D_e"] = 10.0 # Set glucose uptake
medium["EX_o2_e"] = 0.0 # Anaerobic conditions
model.medium = medium
# Calculate minimal media
from cobra.medium import minimal_medium
# Minimize total import flux
min_medium = minimal_medium(model, minimize_components=False)
# Minimize number of components (uses MILP, slower)
min_medium = minimal_medium(
model,
minimize_components=True,
open_exchanges=True
)
7. Flux Sampling
Sample the feasible flux space:
from cobra.sampling import sample
# Sample using OptGP (default, supports parallel processing)
samples = sample(model, n=1000, method="optgp", processes=4)
# Sample using ACHR
samples = sample(model, n=1000, method="achr")
# Validate samples
from cobra.sampling import OptGPSampler
sampler = OptGPSampler(model, processes=4)
sampler.sample(1000)
validation = sampler.validate(sampler.samples)
print(validation.value_counts()) # Should be all 'v' for valid
8. Production Envelopes
Calculate phenotype phase planes:
from cobra.flux_analysis import production_envelope
# Standard production envelope
envelope = production_envelope(
model,
reactions=["EX_glc__D_e", "EX_o2_e"],
objective="EX_ac_e" # Acetate production
)
# With carbon yield
envelope = production_envelope(
model,
reactions=["EX_glc__D_e", "EX_o2_e"],
carbon_sources="EX_glc__D_e"
)
# Visualize (use matplotlib or pandas plotting)
import matplotlib.pyplot as plt
envelope.plot(x="EX_glc__D_e", y="EX_o2_e", kind="scatter")
plt.show()
9. Gapfilling
Add reactions to make models feasible:
from cobra.flux_analysis import gapfill
# Provide a universal reaction database (SBML/JSON); not bundled in cobra 0.31+
from cobra.io import read_sbml_model
universal = read_sbml_model("path/to/universal_reactions.xml")
# Perform gapfilling
with model:
# Remove reactions to create gaps for demonstration
model.remove_reactions([model.reactions.PGI])
# Find reactions needed
solution = gapfill(model, universal)
print(f"Reactions to add: {solution}")
from cobra.io import load_model
# Load model (textbook = fast tutorial; iJO1366 / iML1515 for genome-scale)
model = load_model("textbook")
# Run FBA
solution = model.optimize()
print(f"Growth rate: {solution.objective_value:.3f} /h")
# Show active pathways
print(solution.fluxes[solution.fluxes.abs() > 1e-6])
Workflow 2: Gene Knockout Screen
from cobra.io import load_model
from cobra.flux_analysis import single_gene_deletion
# Load model
model = load_model("textbook")
baseline = model.slim_optimize()
# Perform single gene deletions
results = single_gene_deletion(model)
# Find essential genes (growth < threshold)
essential_genes = results[results["growth"] < 0.01]
print(f"Found {len(essential_genes)} essential genes")
# Find genes with minimal impact
neutral_genes = results[results["growth"] > 0.9 * baseline]
Workflow 3: Media Optimization
from cobra.io import load_model
from cobra.medium import minimal_medium
# Load model
model = load_model("textbook")
# Calculate minimal medium for 50% of max growth
target_growth = model.slim_optimize() * 0.5
min_medium = minimal_medium(
model,
target_growth,
minimize_components=True
)
print(f"Minimal medium components: {len(min_medium)}")
print(min_medium)
Workflow 4: Flux Uncertainty Analysis
from cobra.io import load_model
from cobra.flux_analysis import flux_variability_analysis
from cobra.sampling import sample
# Load model
model = load_model("textbook")
# First check flux ranges at optimality
fva = flux_variability_analysis(model, fraction_of_optimum=1.0)
# For reactions with large ranges, sample to understand distribution
samples = sample(model, n=1000)
# Analyze specific reaction
reaction_id = "PFK"
import matplotlib.pyplot as plt
samples[reaction_id].hist(bins=50)
plt.xlabel(f"Flux through {reaction_id}")
plt.ylabel("Frequency")
plt.show()
Workflow 5: Context Manager for Temporary Changes
Use context managers to make temporary modifications:
# Model remains unchanged outside context
with model:
# Temporarily change objective
model.objective = "ATPM"
# Temporarily modify bounds
model.reactions.EX_glc__D_e.lower_bound = -5.0
# Temporarily knock out genes
model.genes.b0008.knock_out()
# Optimize with changes
solution = model.optimize()
print(f"Modified growth: {solution.objective_value}")
# All changes automatically reverted
solution = model.optimize()
print(f"Original growth: {solution.objective_value}")
Key Concepts
DictList Objects
Models use DictList objects for reactions, metabolites, and genes - behaving like both lists and dictionaries:
# Access by index
first_reaction = model.reactions[0]
# Access by ID
pfk = model.reactions.get_by_id("PFK")
# Query methods
atp_reactions = model.reactions.query("atp")
Flux Constraints
Reaction bounds define feasible flux ranges:
Irreversible: lower_bound = 0, upper_bound > 0
Reversible: lower_bound < 0, upper_bound > 0
Set both bounds simultaneously with .bounds to avoid inconsistencies
Gene-Reaction Rules (GPR)
Boolean logic linking genes to reactions:
# AND logic (both required)
reaction.gene_reaction_rule = "gene1 and gene2"
# OR logic (either sufficient)
reaction.gene_reaction_rule = "gene1 or gene2"
# Complex logic
reaction.gene_reaction_rule = "(gene1 and gene2) or (gene3 and gene4)"
Exchange Reactions
Special reactions representing metabolite import/export:
Named with prefix EX_ by convention
Positive flux = secretion, negative flux = uptake
Managed through model.medium dictionary
Best Practices
Use context managers for temporary modifications to avoid state management issues
Validate models before analysis using model.slim_optimize() to ensure feasibility
Check solution status after optimization - optimal indicates successful solve
Use loopless FVA when thermodynamic feasibility matters
Set fraction_of_optimum appropriately in FVA to explore suboptimal space
Parallelize computationally expensive operations (sampling, double deletions) — start with small n and processes=1 on genome-scale models
Prefer SBML format for model exchange and long-term storage
Use slim_optimize() when only objective value needed for performance
Validate flux samples to ensure numerical stability
Confirm output paths before writing CSV/PNG files from workflow examples
Troubleshooting
Infeasible solutions: Check medium constraints, reaction bounds, and model consistency
Slow optimization: Try different solvers (GLPK, CPLEX, Gurobi) via model.solverUnbounded solutions: Verify exchange reactions have appropriate upper bounds
Import errors: Ensure correct file format and valid SBML identifiers
References
For detailed workflows and API patterns, refer to:
