Skip to main content
. 2020 Dec 15;19:226–246. doi: 10.1016/j.csbj.2020.12.003

Table 4.

Engineering modelling applications. Grouped by the optimization community goal. Focus on optimization/engineering topics. ‘Production’ group includes to optimize different community parameters (strains ratio, carbon source ratio, initial biomass, etc). GR = Growth Rate. Output means the configuration parameters that are predicted. If there is a software available, it is referred to and linked in the column ‘references’ too.

Specific goal of optimization Output Strains Results and additional details Ref.
Production
Maximizing ethanol production
  • -

    carbon source ratio (glucose/xylose)

  • -

    mutant initial biomasses
  • -

    S. cerevisiae (or S. stipitis)

  • -

    E. coli
  • -

    ethanol productivity of ~ 1.08 gr/L/h

  • -

    In vivo experiments to determine kinetics parameters

 [118], [185], [186]
Maximizing flavonoids production
  • -

    carbon sources ratio (glucose/glycerol)

  • -

    strains ratio

 E. coli strains (flavonoid pathway fragmented in 2 strains)
  • -

    Using a scaled-Gaussian model: carbon source ratio of 0:1 (glucose:glycerol), strains ratio of 7:3 (upstream:downstream)

  • -

    Production of flavonoids to 40.7 ± 0.1 mg/L, i.e. a 970-fold improvement

  • -

    Also in vivo experiments to validate the results

 [119]
2 maximization goals:
  • -

    methane production (high community GR)

  • -

    methane yield (low community GR)
  • -

    initial biomasses (strains ratio)

  • -

    flux rates (input and output metabolites)

 - D. vulgaris
- M. maripaludis
- M. barkeri
  • -

    Predicted (max. methane, ATP and biomass yield) and some in vivo data (biomass yield and ATP maintenance)

  • -

    Low biomass yield per strain, vs community goal

  • -

    2 first strains consortium: 0.45 mol. methane/mol. ethanol

  • -

    In vivo validation with literature data from [187]

 [131]
Maximizing yield
  • -

    initial glucose concentration for stable consortia

  • -

    strains ratio

  • -

    uptake glucose and glycerol
  • -

    E. coli: -glucose specialist CV103-‘respirer’

  • -

    acetate specialist CV101-‘fermenter’

  • -

    glycerol specialist CV116
  • -

    In vivo data from [170]. Originally growing in tryptone

  • -

    3 mutants after evolution in-vivo, with different GRs

  • -

    Glucose limited conditions (LTEE)

  • -

    Chemostat model of competition for a simple sugar

  • -

    In silico model predictions for different glucose concentrations

  • -

    >0.0033% of acetate specialist to allow a viable consortium

  • -

    Strain rations: CV101:CV103:CV116 ~= 0.10:0.65:0.025

  • -

    CV103 best takes up the limiting resource glucose, but excretes acetate and glycerol (and/or a closely-related compound, glycerol 3-phosphate)

 [120], [188]
Maximizing (together):
  • -

    community biomass

  • -

    yield per single strain (OptCom fixed goal)
  • -

    strains ratio

  • -

    substrate uptakes

 - D. vulgaris
- M. maripaludis
  • -

    In vivo data from [75]

  • -

    In silico model with OptCom

  • -

    Strain ratio: 2:1 in vivo and 2.28:1 in silico lactate uptake = 48 µM/h

  • -

    formate and hydrogen accumulation = 0

  • -

    Additional in silico predictions: concentration of acetate, methane, CO2 and total biomass

 OptCom [80]
Maximizing (together):
  • -

    community biomass

  • -

    yield per single strain (OptCom fixed goal)
  • -

    strains ratio

  • -

    O2/CO2 ratio
  • -

    Synechococcus spp (SYN)

  • -

    filamentous anoxygenic phototrophs (FAP) related to Chloroflexus and Roseiflexus spp

  • -

    sulphate-reducing bacteria (SRB)
  • -

    In vivo data from [67].

  • -

    In silico model with OptCom

  • -

    Fluxes ratio O2/CO2 reactions: 0.03–0.07

  • -

    Strain ratio: 1:6:1 experimentally, and from 1:5:1 to 3:5:1 with metagenomics data

  • -

    SYN/FAP strain ratio: 1.5–3.5 in vivo and from 7.94 (with O2/CO2 = 0.07) to 20.26 (0.03) in silico

 OptCom [80]
Maximizing (together):
  • -

    community biomass

  • -

    yield per single strain (OptCom fixed goal)
  • -

    strains ratio

  • -

    substrate uptakes
  • -

    C. cellulolyticum

  • -

    D. vulgaris

  • -

    G. sulfurreducens
  • -

    In vivo data from [161].

  • -

    In silico model with OptCom

  • -

    Biomasses: 0.8:0.1:0.13 in vivo and 0.036:0.0045:0.0059 in silico

  • -

    acetate: 2.7 in vivo and 2.48 in silico

    - CO2: 3.3 in vivo and 3.2 in silico

    - Several metabolite fluxes details in Fig.5

 OptCom [80]
Maximizing uranium reduction
  • -

    strains ratio

  • -

    acetate and Fe(III) uptakes
  • -

    S. oneidensis (acetate producer)

  • -

    G. sulfurreducens

  • -

    R. ferrireducens

    Two first ones are uranium reducers
  • -

    In vivo data from [180].

  • -

    In silico model with OptCom

  • -

    Carbon source: lactate = 5 mM

  • -

    In ammonium excess condition ([NH4] = 400 μM)

  • -

    Decrease in the biomass of the uranium-reducing species (SO, GS):

  • -

    Strain ratio max.community biomass: 0.056:0.051:0.055

  • -

    Strain ratio max.uranium reduction: 0.039:0.041:0.056

  • -

    Acetate (GS/RF): 14.9/1.49 when max.uranium reduction

  • -

    Fe(III) (SO/GS/RF): 28.3/110/2.06 when max.uranium reduction

  • -

    Alternative optimization objective in the manuscript

 OptCom [81]
2 cases of study:
  • -

    maximizing butyrate production

  • -

    maximizing atrazine degradation

 Interventions in medium composition or biomass of strains
  • -

    F. prausnitzii and B. adolescentis

    P. aurescens, H. stevensii, Halobacillus sp.
  • -

    In silico model combining GEMs with a Markov Decision Process

  • -

    Predict how to modify the community over time to reach a state of maximum performance

  • -

    Intervention for max. butyrate: inulin increase

  • -

    Intervention for max. atrazine degradation: depending on the microbiome state, increase of the biomass of H.stevensii is often

 MDPbiomeGEM [96]



Pathway distribution
Optimizing metabolite secretion
Secondary goal: medium composition
  • -

    medium composition

  • -

    2 selected strains

  • −

    secreted metabolite

 122 strains (6 from [77]) and 116 from [76] combined in > 6500 different consortia of 2 members
  • -

    In silico framework to design synthetic communities, evaluating which new metabolites could be secreted

  • -

    secreted emergent metabolites (highlighting the most common ones), with their associated two-strain consortium and medium composition

  • -

    E. coli/B. subtilis emergent secretion of both succinate and urea (see Figure S4 and F6 from the original study for more pairs and metabolites)

 [189]
Maximizing growth or compound yield Allocated reactions per strain 2 generic bacteria with reduced central carbon metabolism
  • -

    In silico model following a MILP optimization approach (higher computational cost than LP (FBA)), with a Static/Multi-part method

  • -

    Given metabolic reactions to distribute

  • -

    Strains can only survive through cross-feeding

 [122]
Minimizing number of species Selected species to combine in the community Human gut microbiome
  • -

    In silico model with CoMiDA

  • -

    Graph-based approach (not GEM) combined with Integer Linear Programming (ILP)

  • -

    Given selected substrates and products, and a set of available species

  • -

    Identify putative metabolic pathways from substrates to product

  • -

    Glycolysis pathway, glucose → pyruvate, 284 species: minimal solution with one species was found. Also, they forced for multi-species solution

  • -

    With 10,000 random pairs of substrate-product metabolites, 1–3 species are selected among 2051 species

 CoMiDA [121]
2 cases of study:
  • -

    maximizing antibiotics production,

  • -

    maximizing 1,3-propanediol and methane yield

  • -

    Secondary goal: production

 All reactions to include and their distribution among strains - Streptomyces cattleya and M. barkeri (selected from 4 strains)
- K. pneumoniae and M. mazei
  • -

    In silico model with MultiPlus, following static/Unified approach

  • -

    De novo synthesis of bioactive metabolites

  • -

    Results:

  • -

    Case study 1 (antibiotics): 4 solutions with 528 reactions (2 transports, 3 insertions, and 28 endogenous reactions)

  • -

    Case study 2 (industrial): 6 solutions with 110 reactions (1 transition and 10 endogenous reactions)

 MultiPlus [124]
Optimizing metabolic exchange rates
  • -

    carbon/nitrogen exchange and uptake rates

  • -

    kinetic parameters
  • -

    C. acetobutylicum

  • -

    Wolinella succinogenes
  • -

    In silico model with DMMM, following a dynamic approach

  • -

    Model parameters adjusted to in vivo data (kinetic ones, biomass, carbon and nitrogen sources ratio)

  • -

    Anaerobic species with hydrogen and nitrogen cross-feeding

  • -

    Co-cultures with uni- and multidirectional metabolic interactions

  • -

    The metabolic models can simulate their experimental data, in 4 different cultivation conditions (with/out NH4 and/or NO3), with distinct metabolic capabilities

 [190]
Surviving under constraints Cross-feeding partnerships and division of labor E. coli (2–3 strains)
  • -

    In silico model with DOLMN, following a MILP optimization approach, with a Static/Multi-part method

  • -

    Results:

  • -

    core: 91 combinations of 2 strains. Split the TCA cycle into two halves

  • -

    full with reduced functionalities: 2207 combinations for 2 strains, and 2402 for 3 strains. At least 215 and 203 internal reactions to grow, respectively for 2 and 3 strain consortia. Loss one reaction is not compensated with adding one metabolite in the medium (nonlinear boundary)

 DOLMN [125]
Maximizing ethanol yield KO in strains S. cerevisiae
E. coli
  • -

    In silico model with BioLEGO 2. Based on Microsoft Azure Cloud.

  • -

    Analysis of two-step fermentation pathway of Ulva sp. biomass into ethanol with KOs in each strain from the consortium

  • -

    6,649,115 possible single KO analysed scenarios

  • -

    Ethanol yield increased at 170% of WT (for 867 KO candidate pairs)

 BioLEGO 2 [126]



Stability
Maximizing (together):
  • -

    biomass per single strain

  • -

    community biomass concentration (cells/L)

 strains ratio Auxotrophic E. coli pairs: (argH-lysA)
(lysA, trpC)
(metA, ilvE)
  • -

    In vivo data from [169].

  • -

    In silico model with dOptCom

  • -

    Biomass ratios (approx. values from Fig. 2):

argH-lysA: 0.8:0.2 in vivo and 0.97:0.03 in silicolysA-trpC: 0.9:0.1 in vivo and 0.98:0.02 in silicometA-ilvE: 0.15:0.85 in vivo and 0.15:0.85 in silico 		dOptCom [81]
GR in auxotroph evolution strains ratio E. coli lysine and leucine KOs long-term
  • -

    In vivo data to constrains the model

  • -

    Glucose minimal medium, with uptake rate 10 mmol/gDW/hour

  • -

    Increased GR by 3 folds, while decreased growth in mono-culture

  • -

    Strain ratio depending on the aa uptake rate

 [156]
Common growth
Secondary goal:
spatial distribution
  • -

    strains ratio

  • -

    cross-feeding rate

  • -

    spatial distribution
  • -

    E. coli (KO metE) in lactate

    - S. enterica (secretes methionine)
  • -

    In vivo data from [191] and itself

  • -

    In silico model with COMETS

    Strain ratios: E. coli:S. enterica = 75–80:25–20%

  • -

    Spatial distribution: presence of a strain competitor between cross-feeding species reduces the growth of those strains

 COMETS [89]
GR with optimum distribution of resources
  • -

    metabolites (amino-acids) consumption

 E. rectale or F. prausnitzii, B. thetaiotaomicron, B. adolescentis and R. bromii
  • -

    In vivo data to constrains the model

  • -

    In silico model with CASINO

  • -

    Quantifying diet-induced metabolic changes of the human gut microbiome, using metabolomics data

 CASINO [101]
Common growth
  • -

    strains ratio

  • -

    community GR
  • -

    4 E. coli auxotrophic for amino acids

  • -

    Gut microbiome (9 species)
  • -

    In silico model with SteadyCom

  • -

    4 E. coli case of study:

  • -

    GR: 0.736 gDWh−1

  • -

    Strains ratio: Ec1-Ec2 = 50%, Ec3-Ec4 = 50%. Direct competition Ec1-Ec4 and Ec2-Ec3

  • -

    Gut microbiome case of study: values depending on fibre uptake from B. thetaiotaomicron:

  • -

    GR: ~0.06–0.08 gDWh−1, variable depending on fibre uptake

 SteadyCom [83]



Medium composition
Minimizing the cost of metabolic cooperation Combination of nutrients allowing synergistic growth E. coli arginine and leucine KOs
  • -

    In silico model following a static/Multi-part approach

  • -

    Selected nutrients: supplementation of nucleotide precursors (maltose, xanthine and inosine) to the medium

  • -

    In vivo experimental validation: the predicted medium allows growth

 [127]



Spatial organization
Spatial Partitioning
  • -

    -spatial distribution

  • -

    biofilm thickness

  • -

    growth with by-products

 P. aeruginosa
S. aureus (chronic wound biofilm)
  • -

    In silico dynamic model combining GEM with partial differential equations

  • -

    Results:

  • -

    Tendency of the two bacteria to spatially partition, as observed experimentally. Nutrient gradients influence (oxygen-top-aerobic, glucose-bottom-anaerobic)

  • -

    Different biofilm thickness than isolated

 [129]
Spatial Partitioning
  • -

    -spatial distribution

  • -

    strain ratio

  • -

    shift due to perturbations

 2 case of study (reduced models):
  • -

    E. coli, S. enterica

  • -

    P. putida, P. stutzeri
  • -

    In silico model with IndiMeSH, following a dynamic approach

  • -

    Study of soil habitat

  • -

    Compared to COMETS and experimental data

 IndiMeSH [128]



Flexible
Optimizing PHA accumulated
Secondary goal: production
  • -

    initial biomasses

  • -

    NH4 concentration

  • -

    sucrose secretion rate

 - S. elongatus
- P. putida
  • -

    In silico model with FLYCOP

  • -

    biomasses: 2, 0.2 gr/L

  • -

    NH4: 0.5 mM

  • -

    sucrose secretion rate: 40%

  • -

    PHA production: 22.43 mM/100 h

 FLYCOP [59]
Stability maximization (common growth)
  • -

    strains ratio

  • -

    amino acid secretion rate

 4 E. coli auxotrophic for amino acids
  • -

    In silico model with FLYCOP

  • -

    strains ratio: Ec1 = 35%, Ec2 = 10%, Ec3 = 15%, Ec4 = 40%

  • -

    aa secretion rate (in terms of %GR): Arg = 1.5, Lys = 2, Met = 1.6, Phe = 1

 FLYCOP [59]
Several optimization goals: maximizing yield or biomass or GR, and minimizing time Uptake rates per strain (glucose, acetate, oxygen) 2 E. coli polymorphism:
  • -

    glucose specialist

  • -

    acetate specialist
  • -

    In silico model with FLYCOP

  • -

    In vivo data from Lenski’s experiment (LTEE)

  • -

    Different configurations are predicted depending on the optimization goal. A polymorphism with 2 strains growing is the best configuration under limited oxygen conditions; else only one strain growing

 FLYCOP [59]