Case study I — an E. coli succinate cell factory

Succinate is one of the US-DOE “top-12” bio-based platform chemicals (a precursor to biodegradable plastics, solvents and food additives), and engineering E. coli to overproduce it is a textbook metabolic-engineering problem. The classic strategy (Millard 1996; Vemuri 2002; Sánchez 2005) is: grow anaerobically and delete the competing fermentation routes so carbon is forced through PEP carboxylase / fumarate reductase to succinate.

This case runs that project end-to-end with ov.synbio on the real e_coli_core genome-scale model — from diagnosis → design → thermodynamic check → the enzyme that carries the flux → the DNA to build it.

pip install 'omicverse[synbio]'

import omicverse as ov
ov.plot_set()
m = ov.synbio.load_gem('e_coli_core')          # real BiGG genome-scale model
print(m.id, '|', len(m.reactions), 'reactions,', len(m.genes), 'genes')

🔬 Starting plot initialization...
🧬 Detecting GPU devices…
✅ NVIDIA CUDA GPUs detected: 1
    • [CUDA 0] NVIDIA H100 80GB HBM3
      Memory: 79.1 GB | Compute: 9.0

   ____            _     _    __                  
  / __ \____ ___  (_)___| |  / /__  _____________ 
 / / / / __ `__ \/ / ___/ | / / _ \/ ___/ ___/ _ \ 
/ /_/ / / / / / / / /__ | |/ /  __/ /  (__  )  __/ 
\____/_/ /_/ /_/_/\___/ |___/\___/_/  /____/\___/                                              

🔖 Version: 2.2.1rc1   📚 Tutorials: https://omicverse.readthedocs.io/
✅ plot_set complete.
e_coli_core | 95 reactions, 137 genes

1 — Diagnosis: wild-type E. coli barely secretes succinate

On aerobic glucose the cell maximises growth and makes almost no succinate. The production envelope shows the growth-vs-succinate trade-off: at the wild-type growth optimum, succinate flux is ~0.

print('WT aerobic growth:', round(ov.synbio.fba(m).objective_value, 3), '/h')
ov.synbio.plot_production_envelope(m, 'EX_succ_e')
None

WT aerobic growth: 0.874 /h

2 — The real strategy: go anaerobic, then knock out the competition

Cut off oxygen (EX_o2_e lower bound = 0). Now fermentation dominates, and strain_design proposes the engineering targets. FSEOF ranks over-expression targets — the enzymes that must carry more flux to make succinate; OptKnock (exact bilevel MILP via StrainDesign) proposes gene deletions that couple succinate to growth.

m_an = ov.synbio.load_gem('e_coli_core')
m_an.reactions.EX_o2_e.lower_bound = 0                       # anaerobic
sd = ov.synbio.strain_design(m_an, 'EX_succ_e', method='fseof')
print('over-express (FSEOF):', list(sd.amplify['reaction'].head(5)))
print('growth-coupled knockouts:', list(sd.knockout['reaction'].head(5)))

over-express (FSEOF): ['MDH', 'FRD7', 'SUCCt3', 'FUM', 'NADH16']
growth-coupled knockouts: ['PFL', 'FORt']

The over-expression hits — PPC (PEP carboxylase), FRD7 (fumarate reductase), FUM (fumarase), MDH (malate dehydrogenase) — are exactly the reductive-TCA enzymes that make succinate, and the knockouts hit PFL (pyruvate formate-lyase) and lactate/acetate routes — the real pflB / ldhA deletion strategy. Now confirm with the rigorous MILP:

ok = ov.synbio.strain_design(m_an, 'EX_succ_e', method='optknock',
                             max_knockouts=3, time_limit=120)
print('exact OptKnock knockout designs:')
ok.knockout.head()

exact OptKnock knockout designs:

	knockouts	n_knockouts
0	[GLNS]	1
1	[ICDHyr]	1
2	[RPI]	1
3	[EX_nh4_e]	1
4	[NH4t]	1

3 — Verify the design: does succinate actually go up?

Apply a designed knockout to the anaerobic model and re-solve: knocking out pyruvate formate-lyase (PFL) forces carbon toward succinate — the guaranteed (growth-coupled) succinate flux rises from ~0.

import numpy as np
def coupled_succ(model):
    g = ov.synbio.fba(model).objective_value
    with model as mm:
        bm = [r for r in mm.reactions if r.objective_coefficient != 0][0]
        bm.lower_bound = 0.99 * g
        mm.objective = 'EX_succ_e'; mm.objective.direction = 'min'
        return ov.synbio.fba(mm).objective_value, g

s0, g0 = coupled_succ(m_an)
with m_an as mko:
    mko.reactions.PFL.knock_out()
    s1, g1 = coupled_succ(mko)
print(f'anaerobic WT : succinate {s0:.2f}  (growth {g0:.2f})')
print(f'ΔPFL strain  : succinate {s1:.2f}  (growth {g1:.2f})')

anaerobic WT : succinate 0.00  (growth 0.21)
ΔPFL strain  : succinate 0.67  (growth 0.18)

4 — Is the pathway thermodynamically feasible?

A design is only useful if the thermodynamics allow it. reaction_dg(method='equilibrator') gives real component-contribution ΔG’° (Noor 2013); max_min_driving_force then finds the concentration-optimised driving force of the weakest step of the reductive branch (PEP → OAA → malate → fumarate → succinate).

route = {
    'PPC': {'pep': -1, 'co2': -1, 'h2o': -1, 'oaa': 1, 'pi': 1, 'h': 1},
    'MDH': {'oaa': -1, 'nadh': -1, 'h': -1, 'mal__L': 1, 'nad': 1},
    'FUM': {'mal__L': -1, 'fum': 1, 'h2o': 1},
}
dg0 = {r: ov.synbio.reaction_dg(s, method='equilibrator') for r, s in route.items()}
mdf = ov.synbio.max_min_driving_force(route, dg0, fixed={'h': 1e-7, 'h2o': 1, 'co2': 1e-4})
print('reductive-TCA ΔG\'° (kJ/mol):', {r: round(v, 1) for r, v in dg0.items()})
print('MDF =', round(mdf.mdf, 2), 'kJ/mol | feasible:', mdf.feasible, '| bottleneck:', mdf.bottleneck)
ov.synbio.plot_driving_forces(mdf)
None

reductive-TCA ΔG'° (kJ/mol): {'PPC': -40.3, 'MDH': -26.5, 'FUM': 3.4}
MDF = 14.51 kJ/mol | feasible: True | bottleneck: MDH

5 — The enzyme that carries the flux → back to the model

PEP carboxylase (Ppc) is the entry enzyme of the succinate branch. Its turnover number sets how much protein the cell must invest to carry that flux. Predict k_cat for Ppc with DLKcat (real trained model) on its substrate PEP, push it into an enzyme-constrained model, and see the growth cost — the A↔B hinge.

# real E. coli PEP carboxylase (Ppc) N-terminal fragment + its substrate PEP
PPC = ('MNEQYSALRSNVSMLGKVLGETIKDALGEHILERVETIRKLSKSSRAGNDANRQELLTTLQNLSNDELLPVARAFSQFL'
       'NLANTAEQYHSISPKGEAASNPEVIARTLRKLKNQPELSEDTIKKAVESLSLELVLTAHPTEITRRTLIHKMVEVNACL')
PEP = 'C(=C(/[O-])COP(=O)([O-])[O-])/C(=O)[O-]'
k = ov.synbio.enzyme_kcat(PPC, PEP, method='dlkcat', verbose=False)
print('DLKcat k_cat(Ppc, PEP) =', round(k.kcat, 3), '/s')
ecm = ov.synbio.ec_model(m, {'PPC': k.kcat})
print('growth with Ppc enzyme-constrained:', round(ov.synbio.fba(ecm).objective_value, 3), '/h')

DLKcat k_cat(Ppc, PEP) = 0.835 /s
growth with Ppc enzyme-constrained: 0.489 /h

6 — Build it: codon-optimize the over-expression target + primers

To over-express Ppc you order a synthetic, host-optimised gene and the primers to clone it. codon_optimize rewrites it for E. coli codon usage (GC-bounded, restriction-site-free); design_primers returns validated PCR pairs.

ppc_dna = ov.synbio.codon_optimize(PPC, host='e_coli').sequence
# assemble a ppc over-expression cassette: T7 promoter + RBS + gene + T7 terminator
cassette = ('TAATACGACTCACTATAGG' + 'AAAGGAGGACAACAT' + ppc_dna
            + 'CTAGCATAACCCCTTGGGGCCTCTAAACGGGTCTTGAGGGGTTTTTTG')
print('expression cassette:', len(cassette), 'bp')
ov.synbio.view_construct(cassette, circular=True, title='ppc over-expression plasmid')
None

expression cassette: 556 bp

# PCR primers to clone the ppc gene — shown as binding arrows on the template
primers = ov.synbio.design_primers(ppc_dna)
print('cloning primers:', primers[0])
ov.synbio.view_primers(ppc_dna, primers[0])
None

cloning primers: PrimerPair(product=134bp, Tm=60.0/60.0, L=CCATTAAAGATGCGCTGGGC, R=ATCGTTGCTCAGGTTCTGCA)

Summary

In one notebook we reproduced a real succinate cell-factory design: diagnosed that wild-type E. coli can’t make succinate (production envelope), applied the real anaerobic + knock-out-the-competition strategy (strain_design FSEOF + exact OptKnock → PPC/FRD7 over-expression, PFL/lactate deletion), verified the ΔPFL strain couples succinate to growth, checked the thermodynamics of the reductive-TCA branch with real eQuilibrator ΔG + MDF, connected it to the enzyme carrying the flux (DLKcat k_cat → enzyme-constrained yield), and produced the codon-optimised gene + primers to build it — the full design-build cycle for a metabolic-engineering target, on a real GEM.