代谢网络建模

stcnjtech

已于 2023-10-05 19:56:35 修改

阅读量2.5k

点赞数 33

分类专栏：算法文章标签： python

于 2023-09-08 23:14:10 首次发布

本文链接：https://blog.csdn.net/m0_64163073/article/details/132766899

版权

算法专栏收录该内容

2 篇文章

订阅专栏

1.关于工具包：

平台：matlab和python

1）python平台：

优点：基础功能齐全+版本较新+易用+可视化强

缺点：高级功能不全，需要配合使用不同的工具包，版本兼容复杂，学习成本高+优化不好

（1）cobrapy(官方基本包)

下载链接：https://pypi.org/project/cobra/

学习文档：https://cobrapy.readthedocs.io/en/0.26.3/index.html

注：1.conda+vscode/pycharm

2.利用conda新建一个虚拟环境（python3.7以上）

conda create -n cobraenv python=3.7

3.推荐：在虚拟环境中安装cobrapy0.20.0以上

pip install cobra==0.26.3(最新版)
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple cobra==0.26.3

4.gurobipy：https://pypi.org/project/gurobipy/

pip install gurobipy==10.0.2
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple gurobipy==10.0.2

（2）cameo(提供基本的应变设计，例如OptKnock,OptGene和FSEOF)

下载链接：https://pypi.org/project/cameo/

学习文档： https://cameo.bio/tutorials.html

注：1.conda+vscode/pycharm

2.利用conda新建一个虚拟环境（python3.7）

conda create -n cameoenv python=3.7

3.推荐：在虚拟环境中安装cobrapy0.20.0+cameo0.13.6

pip install cobra==0.20.0(最新版)
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple cobra==0.20.0

pip install cameo==0.13.6
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple cameo==0.13.6

4.gurobipy：https://pypi.org/project/gurobipy/

pip install gurobipy==10.0.2
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple gurobipy==10.0.2

（3）straindesign(提供各种高级应变设计):

下载链接和学习文档：https://straindesign.readthedocs.io/en/latest/index.html

注：1.conda+vscode/pycharm

2.利用conda新建一个虚拟环境（python3.7）

conda create -n straindesignenv python=3.7

3.推荐：在虚拟环境中安装+cobrapy0.26.3+straindesign1.10

pip install cobra==0.26.3(最新版)
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple cobra==0.26.3

pip install straindesign==1.10
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple straindesign==1.10

4.gurobipy：https://pypi.org/project/gurobipy/

pip install gurobipy==10.0.2
或者
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple gurobipy==10.0.2

2）matlab平台(最好配合python平台)：

优点：功能齐全

缺点：版本太老(有些东西报错)+可视化不好+教程不基础

cobra toolbox（只需这个即可）:

下载链接：https://cbb.sjtu.edu.cn/~zwang/sysbio/Lab4COBRA.htm

学习文档：https://opencobra.github.io/cobratoolbox/stable/installation.html

2.model:

必须是BIGG的格式：

网址链接：

http://bigg.ucsd.edu/

https://www.genome.jp/kegg/

3.以下说明基于python平台：

1）FBA：

内容链接：

[cobrapy] https://cobrapy.readthedocs.io/en/0.26.3/simulating.html

[cameo] https://cameo.bio/03-simulate-models.html

[straindesign] https://straindesign.readthedocs.io/en/latest/examples/JN_02_analysis.html

论文参考：

https://www.nature.com/articles/nbt.1614/

FBA Optimize

$maximize/minimize{\ }Z=c^{T}v$
$s.t.\left\{\begin{matrix} Sv = 0&\\ lb\leqslant v_{i}\leqslant ub& \end{matrix}\right.$

FBA algorithm uses a stoichiometric matrix ( $S$ ) of size $m*n$ (m compounds and n reactions) to represent metabolic reactions. The flux through these all of the reactions is represented by a vector $v$ . The objective of FBA is to maximize or minimize an objective function:

$Z=c^Tv$

where $c$ is a vector of weights.

The output of FBA is a particular flux distribution $v$ which is used of linear programming to solve the equation $Sv=0$ given a set of upper and lower bounds on $v$ that maximizes or minimizes the objective function.

For our model STC1515:

$Z=1.0*EX_{15ptnde}-1.0*EX_{15ptndereverse98c21}$

$s.t.{\ }Sv=0$

$v_{i}\in \left\{\begin{matrix} (-10000.0,1000.0) & if{\ } v_{i}{\ }is{\ }reversible\\ (0.0,1000.0) & if{\ } v_{i}{\ }is{\ }not{\ }reversible \end{matrix}\right.$

from pathlib import Path
from cobra.io import load_json_model
from cameo.flux_analysis import fba

# get load path
loadPath = Path(".") / ".."  / "model" / "STC1515.json"
loadPath = loadPath.resolve()
# load model
model=load_json_model(str(loadPath))

# compute FBA algorithm
FBAsolution=fba(model=model,objective=model.reactions.EX_15ptnd_e)

FBAsolution

2）pFBA：

内容链接：

[cobrapy] https://cobrapy.readthedocs.io/en/0.26.3/simulating.html

[cameo] https://cameo.bio/03-simulate-models.html

[straindesign] https://straindesign.readthedocs.io/en/latest/examples/JN_02_analysis.html

论文参考：

https://pubmed.ncbi.nlm.nih.gov/20664636/

from pathlib import Path
from cobra.io import load_json_model
from cameo.flux_analysis import pfba

# get load path
loadPath = Path(".") / ".."  / "model" / "STC1515.json"
loadPath = loadPath.resolve()
# load model
model=load_json_model(str(loadPath))

model.reactions.BIOMASS_Ec_iML1515_core_75p37M

# compute pFBA algorithm
pFBAsolution=pfba(model=model,objective=model.reactions.EX_15ptnd_e)

# get export path
resultPath = Path(".") / ".."  / "result"
resultPath = resultPath.resolve()
# save result as csv
pFBAsolution.data_frame.to_csv(str(resultPath/"pfba.csv"))

3）FVA:

内容链接：

[cobrapy] https://cobrapy.readthedocs.io/en/0.26.3/simulating.html

[cameo] Analyzing models — cameo 0.13.6 documentation

[straindesign] https://straindesign.readthedocs.io/en/latest/examples/JN_02_analysis.html

论文参考：

[1] https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-11-489

[2] https://academic.oup.com/bioinformatics/article/31/13/2159/195895

from pathlib import Path
from cobra.io import load_json_model
from cobra.flux_analysis import flux_variability_analysis

# get load path
loadPath = Path(".") / ".."  / "model" / "STC1515.json"
loadPath = loadPath.resolve()
# load model
model=load_json_model(str(loadPath))

# set objective reaction 
model.objective = 'EX_15ptnd_e'
# compute FVA algorithm
FVAsolution=flux_variability_analysis(model,processes=4)

# get export path
resultPath = Path(".") / ".."  / "result"
resultPath = resultPath.resolve()
# save result as csv
FVAsolution.to_csv(str(resultPath/"fva.csv"))

4）FSEOF：

内容链接：

https://cameo.bio/06-predict-gene-modulation-targets.html

论文参考：

[1] https://pubmed.ncbi.nlm.nih.gov/20348305/

[2] https://www.researchgate.net/figure/Concept-of-FSEOF-framework-to-select-the-gene-amplification-targets-for-enhanced-product_fig1_42611006

[3] https://max.book118.com/html/2019/0215/7020105111002006.shtm

from pathlib import Path
from cameo.io import load_json_model
from cameo.strain_design import FSEOF

# get load path
loadPath = Path(".") / ".."  / "model" / "STC1515.json"
loadPath = loadPath.resolve()
# load model
model=load_json_model(str(loadPath))

fseof = FSEOF(
    model=model,
    # enforced_reaction=model.reactions.EX_15ptnd_e,
    primary_objective=model.reactions.BIOMASS_Ec_iML1515_core_75p37M,
)
fseof_solution=fseof.run(
    target=model.reactions.EX_15ptnd_e,
    max_enforced_flux=0.9,
    number_of_results=20
)

# get export path
resultPath = Path(".") / ".."  / "result"
resultPath = resultPath.resolve()
# save result as csv
fseof_solution.data_frame.to_csv(str(resultPath/'fseof.csv'))

5）OptKnock：

内容链接：

[cameo] https://cameo.bio/05-predict-gene-knockout-strategies.html#optknock

[straindesign] https://straindesign.readthedocs.io/en/latest/examples/JN_06_strain_design_nested.html#OptKnock

论文参考：

https://max.book118.com/html/2019/0215/7020105111002006.shtm

from pathlib import Path
from cameo.io import load_json_model
from cameo.strain_design import OptKnock

# get load path
loadPath = Path(".") / ".."  / "model" / "STC1515.json"
loadPath = loadPath.resolve()
# load model
model=load_json_model(str(loadPath))

# Judgment as exchange reaction (including inlet and outlet)
exchange_reactions = [reaction for reaction in model.reactions if all(s<0 for s in reaction.metabolites.values())]

# Block all outlet reactions
for reaction in exchange_reactions:
    reaction.upper_bound = 0.0
    
# Restore part of the export reaction
model.reactions.EX_15ptnd_e.upper_bound  = 1000.0
model.reactions.EX_ac_e.upper_bound     = 1000.0
model.reactions.EX_co2_e.upper_bound    = 1000.0
model.reactions.EX_etoh_e.upper_bound   = 1000.0
model.reactions.EX_for_e.upper_bound    = 1000.0
model.reactions.EX_h2_e.upper_bound     = 1000.0
model.reactions.EX_h2o2_e.upper_bound   = 1000.0
model.reactions.EX_h2o_e.upper_bound    = 1000.0
model.reactions.EX_h_e.upper_bound      = 1000.0
model.reactions.EX_lac__D_e.upper_bound = 1000.0
model.reactions.EX_meoh_e.upper_bound   = 1000.0
model.reactions.EX_o2_e.upper_bound     = 1000.0
model.reactions.EX_succ_e.upper_bound   = 1000.0
model.reactions.EX_tungs_e.upper_bound  = 1000.0
model.reactions.DM_4crsol_c.upper_bound = 1000.0
model.reactions.DM_5drib_c.upper_bound  = 1000.0
model.reactions.DM_aacald_c.upper_bound = 1000.0
model.reactions.DM_amob_c.upper_bound   = 1000.0
model.reactions.DM_mththf_c.upper_bound = 1000.0
model.reactions.DM_oxam_c.upper_bound   = 1000.0

# Carbon source inlet is set to glucose only
model.reactions.EX_co2_e.lower_bound    = 0.0

# set exclude reactions
exclude_reactions = [reaction.id for reaction in model.reactions if model.genes.s0001 in reaction.genes]
exclude_reactions.append('LYSMO')
exclude_reactions.append('APENTAMAH2')
exclude_reactions.append('APTNAT')
exclude_reactions.append('ARND')
exclude_reactions.append('CARs')
exclude_reactions.append('ARND2')

# build optKnock solver
optknock = OptKnock(
    model=model,
    exclude_reactions=exclude_reactions,
    remove_blocked=True,
    fraction_of_optimum=0.1,
    exclude_non_gene_reactions=True,
    use_nullspace_simplification=True,
)

# run optknock solver
result = optknock.run(
    max_knockouts = 1,
    biomass=model.reactions.BIOMASS_Ec_iML1515_core_75p37M,
    target=model.reactions.EX_15ptnd_e,
    max_results = 1,
)

# check optknock
from cameo.visualization.plotting.with_plotly import PlotlyPlotter
plotter = PlotlyPlotter()
result.plot(plotter, 0)

# get export path
resultPath = Path(".") / ".."  / "result"
resultPath = resultPath.resolve()
# save result as csv
result.data_frame.to_csv(str(resultPath/"optknock.csv"))