- Requirements
- Sequences
- Tools models
- Merlin models
- Reactions identifier conversion
- Drafts assessments
- pip install matplotlib
- pip install biopython
- pip install seaborn
- pip install json
- pip install cobra
- pip install pandas
The gene sequences used to reconstruct each of the Genome-Scale Metabolic Models (GSMM) are stored in the folder "Sequences". The files are in FASTA and GFF format. The FASTA file of the gene sequences used for AuReMe template models were preprocessed to include only the locus tag. The preprocessing scripts and resultant FASTA files are stored in the "Template models genomes processing" folder. This preprocess will convert a file from this:
>lcl|NC_000913.3_prot_NP_414544.1_3 [gene=thrB] [locus_tag=b0003] [db_xref=UniProtKB/Swiss-Prot:P00547] [protein=homoserine kinase] [protein_id=NP_414544.1] [location=2801..3733] [gbkey=CDS]
MVKVYAPASSANMSVGFDVLGAAVTPVDGALLGDVVTVEAAETFSLNNLGRFADKLPSEPRENIVYQCWE
RFCQELGKQIPVAMTLEKNMPIGSGLGSSACSVVAALMAMNEHCGKPLNDTRLLALMGELEGRISGSIHY
DNVAPCFLGGMQLMIEENDIISQQVPGFDEWLWVLAYPGIKVSTAEARAILPAQYRRQDCIAHGRHLAGF
IHACYSRQPELAAKLMKDVIAEPYRERLLPGFRQARQAVAEIGAVASGISGSGPTLFALCDKPETAQRVA
DWLGKNYLQNQEGFVHICRLDTAGARVLEN
to this:
>b0003
MVKVYAPASSANMSVGFDVLGAAVTPVDGALLGDVVTVEAAETFSLNNLGRFADKLPSEPRENIVYQCWE
RFCQELGKQIPVAMTLEKNMPIGSGLGSSACSVVAALMAMNEHCGKPLNDTRLLALMGELEGRISGSIHY
DNVAPCFLGGMQLMIEENDIISQQVPGFDEWLWVLAYPGIKVSTAEARAILPAQYRRQDCIAHGRHLAGF
IHACYSRQPELAAKLMKDVIAEPYRERLLPGFRQARQAVAEIGAVASGISGSGPTLFALCDKPETAQRVA
DWLGKNYLQNQEGFVHICRLDTAGARVLEN
This conversion is performed so that AuReMe could map the locus tag of each gene to the genes found in the template GSMM.
The genome of Escherichia coli was obtained from the NCBI assembly NC_000913.3. The genome of Lactococcus lactis was obtained from the NCBI assembly NC_009004.1. The genome of Plasmodium falciparum was obtained in https://plasmodb.org/plasmo/app as in https://doi.org/10.1371/journal.pcbi.1005895.
The genome files used for the reconstruction of the models of Bordetella pertussis and Lactobacillus plantarum were obtained from here and here.
The genome files used for the reconstruction of the models of Toxoplasma gondii were obtained in here as in https://doi.org/10.1371/journal.pcbi.1004261.
Each draft and curated model is stored in the "Models" folder.
AuReMe (Automatic Reconstruction of Metabolic Models) is a Docker-based software that allows reconstructing simulation-ready GSMMs using others as templates or inheriting a genome annotation from Pathway Tools.
The AuReME Docker image version 2.4 was utilised to generate the draft models. AuReMe requires other models as templates and the genome files of the organisms under study, including the template ones. In this sense, the genome FASTA files were preprocessed only to contain the locus tag in the header of each gene. Moreover, the template models were imported to merlin in SBML level 3 and exported in SBML level 2 format. Otherwise, AuReMe could not be run as it does not support SBML level 3 template models. Then, an orthology-based reconstruction was generated using the Ortho-Finder algorithm.
Pathway Tools is software designed for the development of organism-specific metabolic databases. The application is developed in Lisp and provides a graphical interface that allows the user to visualise the whole constructed network. Moreover, it includes the assisted analysis of experimental data in the metabolic network.
The Pathway Tools version 25.5 was downloaded and utilised to generate draft models of the mentioned organisms under study. For this matter, we uploaded the genome file in GeneBank format and performed the metabolic network reconstruction with the Pathologic algorithm using default parameters. Moreover, transport reactions were generated and included. Finally, the draft model was exported in an SBML format.
CarveMe is a python package focused on fast reconstructing microbial metabolic models. It provides a top-down approach that creates simulation-ready models based on BiGG universal database. This tool uses Diamond to prioritise reactions with stronger genetic evidence (by similarity).
The protein sequences were used as input in FASTA format using default parameters, providing simulation-ready GSM models in SBML format. The gap-filling mode of this tool was not used.
RAVEN (Reconstruction, Analysis and Visualization of Metabolic Networks) is a command-line tool compliant with COBRA Toolbox v3, thus running in MATLAB. It provides functionalities to reconstruct models from scratch or use template models, merge networks from different databases and curate from the command line.
The protein FASTA files were used as input to the version 2.5.3 of the tool, using the KEGG’s “prok90_kegg94” and “euk90_kegg94” templates for prokaryotic and eukaryotic organisms, respectively. The tool was used with default parameters, except for the “keepUndefinedStoich” and “keepIncomplete” (were set as False), and the cutOff value (adjusted to 10-30).
ModelSEED is a web tool to analyse and reconstruct GSMMs. The genome annotation is performed by RAST within ModelSEED, reconstructing the metabolic network based on that annotation. Moreover, it provides a graphical interface to analyse and gap-fill the model as well as run simulations.
ModelSEED version 2 (November 2021) was used to reconstruct the draft models of the organisms under study. The genome FASTA files were inputted directly through the web interface, and the template models were defined by each type of organism (gram-positive bacteria for L. plantarum and gram-negative bacteria for B. pertussis). Finally, we selected a complete medium for the reconstruction.
AutoKEGGRec v1.01 is a command-line tool that runs in MATLAB and provides compliance with COBRA Toolbox v3. This tool is highly suitable to generate multiple GSMMs only in one run and can also generate metabolic models for communities. The reconstruction is dependent on the availability of data for the target organism in KEGG.
The KEGG’s organism identifiers for L. plantarum, B. pertussis, and T. gondii were used as input, using the “SingleRecs” mode, suitable for reconstruction of GSM models of individual organisms. After running the tool, the models were exported in SBML format.
Merlin BIT models were generated on the “use specific BiGG models information” option where we selected the template models.
Merlin models will be referred to from hereafter using the following nomenclature:
-
Merlin-DFA – merlin model using Diamond “fast” mode and Automatic workflow;
-
Merlin-DFS – merlin model using Diamond “fast” mode and SamPler;
-
Merlin-DSA – merlin model using Diamond “sensitive” mode and Automatic workflow;
-
Merlin-DSS – merlin model using Diamond “sensitive” mode and SamPler;
-
Merlin-BA – merlin model using BLAST and Automatic workflow;
-
Merlin-BS – merlin model using BLAST and SamPler;
The phyllogenetic trees results used to assess the organisms for Automatic workflow are stored in the "Phyllogenetic tree" folder.
The SamPler parameters are stored in the "Sampler Parameters" folder.
All the workspaces are available here.
The reactions identifier conversion was performed using the MetaNetX file of cross-references obtained here and downloaded clicking in this link: https://www.metanetx.org/cgi-bin/mnxget/mnxref/reac_xref.tsv.
The obtained TSV format file was preprocessed to exclude the initial lines that refer the type of data, license and so on. Then, reactions without conversion were removed and the file format was converted from
biggR:R_Htmi MNXR03 secondary/obsolete/fantasy identifier
biggR:R_Htr MNXR03 secondary/obsolete/fantasy identifier
biggR:R_Hts MNXR03 secondary/obsolete/fantasy identifier
biggR:R_Htu MNXR03 secondary/obsolete/fantasy identifier
biggR:R_Htx MNXR03 secondary/obsolete/fantasy identifier
biggR:R_PCXHtpp MNXR03 secondary/obsolete/fantasy identifier
metacyc.reaction:RXN-13519 MNXR03 vectoral H+ export||4 metacycM:PROTON@cco:CCO-IN --> 4 metacycM:PROTON@cco:CCO-OUT
to
Internal ID,External ID,Source
MNXR03,R_Htmi,BiGG
MNXR03,R_Htr,BiGG
MNXR03,R_Hts,BiGG
MNXR03,R_Htu,BiGG
MNXR03,R_Htx,BiGG
MNXR03,R_PCXHtpp,BiGG
MNXR03,RXN-13519,MetaCyc
This format conversion can be performed using the jupyter notebook "Scripts/Xrefs files/xrefs_files_reducer.ipynb".
The comparison of the reaction sets between the draft and curated models considered the following:
-
True Positives (TP) - reactions of the draft model with at least one alias in the curated model.
-
False Positives (FP) - reactions of the draft model without any alias in the curated model.
-
False Negatives (FN) - absent reactions in the draft model but present in the curated model.
The comparison of the gene sets between the draft and curated models considered the following:
-
True Positives (TP) - genes present in the draft and curated models.
-
False Positives (FP) - genes present in the draft models and absent in the curated model.
-
False Negatives (FN) - absent genes in the draft models but present in the curated model.
The scripts for draft models assessment are stored in the "Scripts" folder. The models assessment is present in the "Scripts/assessment.ipynb" jupyter notebook. The graphs can be generated as in "Scripts/graphs_generation.ipynb". The results of this assessment are stored in the "Results" folder.