MCS extended

An extended and generalized MCS framework

2020/06/15

Philipp Schneider, Axel von Kamp, Steffen Klamt

Added Features:

1. Definition of multiple target regions (T·r ≤ t) and multiple desired regions (D·r ≤ d).

2. Specification of gene and reaction deletion and addition candidates and individual cost factors for each intervention.

3. Fast computation of gene-based MCS using GPR associations and novel compression techniques.

Software Requirements:

1. MATLAB 2016b® or later

2. IBM ILOG® CPLEX® 12.7, 12.8, 12.9 or 12.10 (Make sure to use compatible CPLEX® and MATLAB® versions. Version 12.10 is recommended.)

3. CellNetAnalyzer2020.2

4. Set up CellNetAnalyzer to access the CPLEX-Matlab-API (replace the CPLEX-default paths in startcna.m with the paths to your CPLEX installation as described by CellNetAnalyzer manual)

Getting Started:

1. Download this project to your computer (see release page https://github.com/ARB-Lab/MCS_extended/releases) and extract all files.
2. Start MATLAB.
3. Navigate to the main directory of your installation of the CellNetAnalyzer toolbox. By executing pwd you can verify that the CNA main directory is now your current woking directory.
4. Start CellNetAnalyzer by executing startcna or startcna(1) (silent start).
5. Either navigate to the path of the script in this project that you want to execute or add all folders from this project to your MATLAB® path by a right click on the project folder and "Add to path" -> "Selected folders and subfolders"

Script Files:

1. cofeeding_example/cofeeding_example.m

   Demonstrates how CNAMCSEnumerator2 can be used to compute strain designs for single subtrate feeding and substrate co-feeding from a single MCS setup.

2. GPR_example/GPR-example.m

   Demonstrates how CNAgeneMCSEnumerator2 computes gene-based MCS using GPR associations and advanced compression routines.

3. e_coli/benchmark.m

   Computes and characterizes gene-MCS for the production of 2,3-BDO in E. coli in a core (ECC2) and a genome-scale (iML1515) setup. The scipt benchmarks the runtime reduction achieved through applying the novel GPR rule compression. The computed MCS for the genome-scale setup (scenario 1) also serve as a reference for scripts 4-6. Results are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement. By changin variables in the script to e.g. model='full', options.compression_GPR or options.preproc_compression, different setups and compression routines can be used.

   model='ECC2': Computation from an E. coli core model with ca. 500 reactions model='full': Computation from a genome-scale E. coli model (iML1515) with ca. 2700 reactions

4. e_coli/desired2.m

   Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in E.coli from a similar setup as in scenario 1 (benchmark.m). A second desired region is added to the setup to ensure that strain designs support higher ATP maintanance rates (scenario 2 in Table2). The results of scenario 2 are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

5. e_coli/des1tar2.m

   Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in E.coli. A second target region is added to scenario 1 to compute, at the same time, single substrate and co-feeding strategies using glucose, acetate and glycerol (scenario 3 in Table 2). Therefore, the supply reactions for glucose, acetate and glycerol are specified as addition candidates. Results of scenario 3 are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

6. e_coli/des2tar2.m

   Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in E. coli. In addition to scenario 3, a second desired region is added to demand the support of higher ATP maintanance rates (scenario 4 in Table 2). This setup shows that a combination of multiple target and desired regions is possible and generates again qualitatively new solutions. Results of scenario 4 are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

7. pseudomonas/pseudomonas_benchmark.m

   Computes and characterizes gene-MCS for the production of 2,3-BDO in Pseudomonas putida from Glucose. By changing options.compression_GPR or options.preproc_compression, GPR- and/or network compression can be activated or deactivated. Results are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

8. pseudomonas/pseudomonas_desired2.m

   Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in P. putida from a similar setup as in pseudomonas_benchmark. A second desired region is added to the setup to ensure that strain designs support higher ATP maintanance rates. The results of this computation are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

9. pseudomonas/pseudomonas_des1tar2.m

   Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in P. putida A second target region is added to the benchmark setup to compute, at the same time, single substrate and co-feeding strategies using glucose, acetate and glycerol. Therefore, the supply reactions for glucose, acetate and glycerol are specified as addition candidates. Results of this computation are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

10. pseudomonas/pseudomonas_des2tar2.m

    Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in P. putida In addition to the changes in pseudomonas_des1tar2, a second desired region is added to demand the support of higher ATP maintanance rates. This setup shows that a combination of multiple target and desired regions is possible and generates again qualitatively new solutions. Results of this computation are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table.

11. yeast/yeast_benchmark.m

    Computes and characterizes gene-MCS for the production of 2,3-BDO in Saccharomyces cerevisiae from Glucose. By changing options.compression_GPR or options.preproc_compression, GPR- and/or network compression can be activated or deactivated. Results are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

12. yeast/yeast_desired2.m

    Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in S. cerevisiae from a similar setup as in yeast_benchmark. A second desired region is added to the setup to ensure that strain designs support higher ATP maintanance rates. The results of this computation are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

13. yeast/yeast_des1tar2.m

    Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in S. cerevisiae A second target region is added to the benchmark setup to compute, at the same time, single substrate and co-feeding strategies using glucose and acetate. Therefore, the supply reactions for glucose and acetate are specified as addition candidates. Results of this computation are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

14. yeast/yeast_des2tar2.m

    Computes and characterizes genome-scale gene-MCS for the production of 2,3-BDO in S. cerevisiae In addition to the changes in yeast_des1tar2, a second desired region is added to demand the support of higher ATP maintanance rates. This setup shows that a combination of multiple target and desired regions is possible and generates again qualitatively new solutions. Results of this computation are saved to a .mat file in the working directory and the MCS characterization/ranking is saved as a .tsv table. By default MCS characterization is skipped because of a return statement after MCS computation. To activate it, remove this return statement.

15. synthetic_lethals/synthetic_lethals_iML.m

    Computes synthetic lethals in the genome scale E. coli model iML1515 up to the size of 4 gene knockouts via the gMCS algorithm of Apaolaza and via our MCS algorithm in CellNetAnalyzer. The results of both functions are checked for consistency and the runtimes of both methods are returned in a command window output.

Minor functions required for the scripts above:

16. functions/verify_mcs.m

17. functions/cell2csv.m

18. functions/relev_indc_and_mdf_Param.m

19. functions/start_parallel_pool_on_SLURM_node.m

20. functions/block_non_standard_products

21. functions/compare_mcs_sets.m

22. functions/text2num_mcs.m

23. functions/verify_lethals.m

Model files:

24. e_coli/benchmark_iJOcore.mat - E. coli core model required for script (3)

25. e_coli/iML1515.mat - genome scale E. coli model required for scripts (3-6)

26. pseudomonas/iJN746.mat - genome scale P. putida model required for scripts (7-10)

27. yeast/yeastGEM.xml - genome scale S. cerevisiae model required for scripts (11-14)

28. yeast/yeast_BiGGmetDictionary.csv - Dictionary to replace the species identifiers from yeastGEM with Bigg identifiers

29. yeast/yeast_BiGGrxnDictionary.csv - Dictionary to replace the reaction identifiers from yeastGEM with Bigg identifiers

Relevant new (API) functions included in the most recent release (2020.2) of the CellNetAnalyzer toolbox :

• CNAgeneMCSEnumerator2

  Function wrapper for CNAMCSEnumerator2 that allows the computation of gene-MCS using GPR association and GPR-rule compression, multiple target and desired regions and gene- and reaction deletions and additions with individual intervention cost factors.

• CNAMCSEnumerator2

  MCS computation with multiple target and desired regions, reaction additions and deletions and individual cost factors.

• CNAgenerateGPRrules.m

  Translates GPR-rules provided in text form into a gene-protein-reaction mapping.

• CNAintegrateGPRrules.m

  Extends a metabolic network model with genes and GPR rules represented by pseudoreaction and pseudometabolites. Uses mapping generated by CNAgenerateGPRrules.m

• CNAcharacterizeGeneMCS.m

  Characterizes and ranks geneMCS by different criteria, such as product yield, ability to grow, implementation effort

• testRegionFeas.m

  Tests if a model/mutant has feasible steady state flux vectors in a flux space spanned by a set of constraints (V·r ≤ v).

Remarks:

• If a fast but incomplete iterative MCS computation/search is preferred over a full MCS enumeration, set the parameter "enum_method" (e.g. in scripts 3-6) from 2 to 1 and set a time or solution limit.