This repository contains all source code in the Reconstructor Python package, important file dependencies, and benchmarking scores for Reconstructor models. Reconstructor is a COBRApy compatible, automated GENRE building tool from annotated aminoa acid .fasta files based on KEGG annotations.

Reconstructor is currently only compatible on MacOSX and Windows machines

/RepresentativeGENRES/Reconstructor: Contains 10 representative bacterial GENRES in .sbml format created by Reconstructor from annotated .fasta files.

/RepresentativeGENREs/ModelSEED: Corresponding GENREs created with ModelSEED using the same genome sequence information as used in Reconstructor.

/RepresentativeGRENREs/CarveME: Corresponding GENREs created with CarveME using genome sequences from the same species (or in one case, genus), but not exactly the same strain. These models were taken from the CarveME database (https://github.com/cdanielmachado/carveme).

Contains raw .html files for benchmarking scores for 10 representative Reconstructor models

See below in the Supplementary Model Checks and Analyses at the bottom of the README for links to stable .html renderings of the MEMOTE scores and comparisons to other automatic model generation tools

contains all package source code

In terminal....
Install homebrew if you do not already have it installed by copying/pasting the following line into terminal (further instructions here):

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Download DIAMOND sequence aligner (Version >= 2.0.15) using homebrew:

brew install diamond

Install Reconstructor package using pip:

pip install reconstructor

Use following command to download additional dependencies/run final installation checks (should take ~1hr to complete depending on your system):

python -m reconstructor --test yes

Install Reconstructor package using CMD.exe prompt launched from Anaconda navigator:

pip install reconstructor

Use following command to download additional dependencies/run final installation checks (hould take ~1hr to run depending on your system):

python -m reconstructor --test yes

You must first have the DIAMOND sequence aligner downloaded (MUST BE >= VERSION 2.0.15), installation instructions can be found here: https://github.com/bbuchfink/diamond
If you think you already have DIAMOND installed, you can check your version using the terminal command:

diamond --version

You can also download DIAMOND through homebrew (if you have homebrew installed) using the terminal command:

brew install diamond

You do not need to install DIAMOND on your windows machine. A Windows executable function is already pre-packaged within the Reconstructor software.

This can be done via pip in terminal on Mac, or a CMD.exe Prompt launched from Anaconda Navigator in Widnows:

pip install reconstructor

You must be running >= Python 3.8
To determine your Python version you can use the following command:

python --version

Run the following test to ensure reconstruction was installed correctly and is functional, and downloads database files that are necessary for reconstructor to work. This series of tests should take about an hour to run, dependent on computer/processor speed. These are runtimes for Reconstructor on a 2020 MacBook Pro with a 1.4 GHz Quad-Core Intel Core i5 processor.

MAC USERS MAY BE ASKED FOR TERMINAL TO HAVE ACCESS TO DOWNLOADS, CAMERA, LOCATION, ETC. Please allow terminal to have access to all locations on your computer. Reconstructor will NOT gather data from your camera, location, or other sensitive information. Reconstructor is simply searching for the file titled glpk_interface.py on your local machine (installed when COBRA module is installed) and replacing it with a newer, functional version.

Use the command below to test reconstructor to ensure correct installation:

python -m reconstructor --test yes

*YOU MUST RUN THE TEST SUITE BEFORE PROCEEDING TO USE RECONSTRUCTOR

Now that Reconstructor and all dependency databases are installed, you can proceed to use Reconstructor via command line. An example would be:

python -m reconstructor --input_file <input fasta file> --file_type <1,2,3> --gram <negative, positive> --media rich

python -m reconstructor --input_file Osplanchnicus.aa.fasta --file_type 1 --gram negative --media rich

python -m reconstructor --input_file Osplanchnicus.hits.out --file_type 2 --gram negative --media rich

python -m reconstructor --input Osplanchnicus.sbml --type 3 --media rich

You can use Reconstructor directly in Python for using directly with COBRApy analysis tools. To import the reconstruction function use the following line:

from reconstructor import reconstruct

The reconstruct function is defined as follows, with the required argument being input_file. All other arguments have default options following the equals sign. Argument descriptions are provided below

reconstruct(input_file, file_type = 1, media=[], org = 'default', min_frac = 0.01, max_frac = 0.5, gram='none', out = 'default', name = 'default', cpu = 1, gapfill = 'yes')

Here is an example of how to generate a GENRE from an annotated amino acid fasta file (.fa, type 1 input) directly in your python script.

model = reconstruct('218496.4.fa', file_type = 1, gram = 'negative')

--input_file <REQUIRED input file, Required, str>

--file_type <REQUIRED input file type, .fasta = 1, diamond blastp output = 2, .sbml = 3, Required, Default = 1, int> 

--gram <REQUIRED Type of Gram classificiation (positive or negative), default = positive, str>

--media <REQUIRED 'rich' is the default and can be used to generate model based on a rich media. List of strings of metabolites in modelseed namespace composing the media condition, comma separated. Must end with _e. For example: 'cpd00001_e'.>

--org <KEGG organism code. Not required, str>

--min_frac <Minimum objective fraction required during gapfilling, default = 0.01, float>

--max_frac <Maximum objective fraction allowed during gapfilling, default = 0.5 ,float>

--out <Name of output GENRE file, default = default, str>

--name <ID of output GENRE, default = default, str>

--cpu <Number of processors to use, default = 1, int>

--test <run installation tests, default = no, str>

To view rendered .html MEMOTE benchmarking scores for Reconstructor models and the corresponding ModelSEED/CarveME models, use the links below:
Porphyromonas gingivalis: Resonstructor || ModelSEED || CarveME
Bacillus amyloliquefaciens: Resonstructor || ModelSEED || CarveME
Citrobacter braakii: Resonstructor || ModelSEED || CarveME
Acinetobacter bereziniae: Resonstructor || ModelSEED || CarveME
Tropheryma whipplei: Resonstructor || ModelSEED || CarveME
Proteus mirabilis: Resonstructor || ModelSEED || CarveME
Clostridiodes difficile: Resonstructor || ModelSEED || CarveME
Campylobacter jejuni: Resonstructor || ModelSEED || CarveME
Helicobacter pylori: Resonstructor || ModelSEED || CarveME
Escherichia coli: Resonstructor || ModelSEED || CarveME

This folder contains .sbml reconstrucitons for C. difficile on five different media conditions described below. These include a complete media (default reconstructor media if no media is defined), rich media (built into reconstructor), minimal media (built into reconstructor), C. difficile specific defined minimal media (user inputted), and a C. difficile specific defined enriched media (user inputted).

Default media contains the following ModelSEED compounds:
cpd00035 (L-Alanine), cpd00051 (L-Arginine), cpd00132 (L-Asparagine), cpd00041 (L-Aspartate), cpd00084 (L-Cysteine), cpd00053 (L-Glutamine), cpd00023 (L-Glutamate), cpd00033 (Glycine), cpd00119 (L-Histidine), cpd00322 (L-Isoleucine), cpd00107 (L-Leucine), cpd00039 (L-Lysine), cpd00060 (L-Methionine), cpd00066 (L-Phenylalanine), cpd00129 (L-Proline), cpd00054 (L-Serine), cpd00161 (L-Threonine), cpd00065 (L-Tryptophan), cpd00069 (L-Tyrosine), cpd00156 (L-Valine), cpd00027 (D-Glucose), cpd00149 (Cobalt), cpd00030 (Manganese), cpd00254 (Magnesium), cpd00971 (Sodium), cpd00063 (Calcium), cpd10515 (Iron), cpd00205 (Potassium), cpd00099 (Choride)
This reconstruction was generated using the following arugments:

--input 699034.5.fa --type 1 --gram positive  

The MEMOTE scores for C. difficile reconstruction in complete media can be found here

Rich media contains the following ModelSEED compounds:
cpd00001 (Water), cpd00035 (L-Alanine), cpd00041(L-Aspartate), cpd00023 (L-Glutamate), cpd00119 (L-Histidine), cpd00107 (L-Leucine), cpd00060 (L-Methionine), cpd00161 (L-Threonine), cpd00069 (L-Tyrosine), cpd00084 (L-Cysteine), cpd00033 (Glycine), cpd00322 (L-Isoleucine acid), cpd00066 (L-Phenylalanine), cpd00054 (L-Serine), cpd00065 (L-Tryptophan), cpd00156 (L-Valine), cpd00220 (Riboflavin), cpd00644 (Pantothenate), cpd00393 (Folate), cpd00133 (Nicotinamide), cpd00263 (Pyridoxal), cpd00104 (Biotin), cpd00149 (Cobalt), cpd00971 (Sodium), cpd00099 (Chloride), cpd00205 (Potassium), cpd00009 (Phosphate), cpd00063 (Calcium), cpd00254 (Magnesium), cpd10515 (Iron), cpd00030 (Manganese), cpd00242 (Bicarbonate), cpd00226 (Hypoxanthine), cpd01242 (Thyminose), cpd00307 (Cytosine), cpd00092 (Uracil), cpd00117 (D-Alanine), cpd00067 (Hydrogen), cpd00567 (D-Proline), cpd00132 (L-Asparagine), cpd00210 (Taurine), cpd00320 (D-Aspartate), cpd03279 (Deoxyinosine) , cpd00246 (Inosine), cpd00311 (Guanosine), cpd00367 (Cytidine), cpd00277 (Deoxyguanosine), cpd00182 (Adenosine), cpd00654 (Deoxycytidine), cpd00412 (Deoxyuridine), cpd00438 (Deoxyadenosine), cpd00274 (Citrulline), cpd00186 (D-Glutamate), cpd00637 (D-Methionine), cpd00105 (D-Ribose), cpd00305 (Thiamin), cpd00309 (Xanthine), cpd00098 (Choline), cpd00207 (Guanine), cpd00082 (D-Fructose), cpd00129 (L-Proline) This reconstruction was generated using the following arugments:

--input 699034.5.fa --type 1 --gram positive --media rich_media  

The MEMOTE scores for C. difficile reconstruction in rich media can be found here

Minimal media contains the following ModelSEED compounds:
cpd00001 (Water), cpd00065 (L-Tryptophan), cpd00060 (Methionine), cpd00322 (L-Isoleucine), cpd00129 (L-Proline), cpd00156 (L-Valine), cpd00107 (L-Leucine), cpd00084 (L-Cysteine), cpd00149 (Cobalt), cpd00099 (Chloride), cpd10515 (Iron), cpd00030 (Manganese), cpd00254 (Magnesium), cpd00063 (Calcium), cpd00205 (Potassium), cpd00009 (Phosphate), cpd00971 (Sodium), cpd00242 (Carbonate), cpd00104 (Biotin), cpd00644 (Pantothenate), cpd00263 (Pyridoxine) , cpd00082 (D-Fructose) This reconstruction was generated using the following arugments:

--input 699034.5.fa --type 1 --gram positive --media minimal_media  

The MEMOTE scores for C. difficile reconstruction in minimal media can be found here

C. difficile defined minimal media contains the following ModelSEED compounds:
cpd00001 (Water), cpd00104 (Biotin), cpd00644 (Pantothenate), cpd00263 (Pyridoxine), cpd00149 (Cobalt), cpd00099 (Chloride), cpd10515 (Iron), cpd00030 (Manganese), cpd00254 (Magnesium), cpd00063 (Calcium), cpd00205 (Potassium), cpd00009 (Phosphate), cpd00971 (Sodium), cpd00242 (Carbonate), cpd00322 (L-Isoleucine), cpd00129 (L-Proline), cpd00156 (L-Valine), cpd00107 (L-Leucine), cpd00084 (L-Cysteine), cpd00065 (L-Tryptophan), cpd00027 (Glucose)
This reconstruction was generated using the following arguments:

--input 699034.5.fa --type 1 --gram positive --media 'cpd00001_e','cpd00104_e','cpd00644_e','cpd00263_e','cpd00149_e','cpd00099_e','cpd10515_e','cpd00030_e','cpd00254_e','cpd00063_e','cpd00205_e','cpd00009_e','cpd00971_e','cpd00242_e','cpd00322_e','cpd00129_e','cpd00156_e','cpd00107_e','cpd00084_e','cpd00065_e','cpd00027_e'

The MEMOTE scores for C. difficile reconstruction in minimal defined media can be found here

C. difficile defined enriched media contains the following ModelSEED compounds:
cpd00035 (Alanine), cpd00041 (Aspartic Acid), cpd00023 (Glutamic Acid), cpd00119 (Histidine), cpd00107 (Leucine), cpd00060 (Methionine), cpd00129 (Proline), cpd00161 (Threonine), cpd00051 (Arginine), cpd00069 (Tyrosine), cpd00084 (Cysteine), cpd00033 (Glycine), cpd00322 (Isoleucine), cpd00039 (Lysine), cpd00066 (Phenylalanine), cpd00054 (Serine), cpd00065 (Tryptophan), cpd00156 (Valine), cpd00027 (Glucose), cpd00220 (Riboflavin), cpd00644 (Calcium Pantothentate), cpd00393 (Folate), cpd00133 (Niacin), cpd00263 (Pyridoxine), cpd00104 (Biotin), cpd00149 (Cobalt), cpd00971 (Sodium), cpd00099 (Chloride), cpd00205 (Potassium), cpd00009 (Phosphate), cpd00063 (Calcium), cpd00254 (Magnesium), cpd10515 (Fe2+), cpd00030 (Mn2+), cpd00242 (Carbonate), cpd00001 (Water), cpd00226 (Hypoxanthine), cpd01242 (Thyminose), cpd00307 (Cytosine), cpd00092 (Uracil)
This reconstruction was generated using the following arguments"

--input 699034.5.fa --type 1 --gram positive --media 'cpd00035_e','cpd00041_e','cpd00023_e','cpd00119_e','cpd00107_e','cpd00060_e','cpd00129_e','cpd00161_e','cpd00051_e','cpd00069_e','cpd00084_e','cpd00033_e','cpd00322_e','cpd00039_e','cpd00066_e','cpd00054_e','cpd00065_e','cpd00156_e','cpd00027_e','cpd00220_e','cpd00644_e','cpd00393_e','cpd00133_e','cpd00263_e','cpd00104_e','cpd00149_e','cpd00971_e','cpd00099_e','cpd00205_e','cpd00009_e','cpd00063_e','cpd00254_e','cpd10515_e','cpd00030_e','cpd00242_e','cpd00001_e','cpd00226_e','cpd01242_e','cpd00307_e','cpd00092_e'

The MEMOTE scores for C. difficile reconstruction in enriched defined media can be found here

The universal reaction database used in reconstructor is a modified version of the ModelSEED biochemistry database avaliable here. We generated the universal database using the UniversalCreation.py script. We corrected poorly defined metabolite formulas and removed mass imbalanced reactions using the curateuniversal.py script that is avaliable in this repository.

If you wish to modify the provided universal reaction database you must first locate the universal.pickle file that was downloaded during the reconstructor installation phase on your local computer. You can modify this file by running a python script. You must write this script yourself or use curateuniversal.py as a template. You can then run this script in your command line.

Begin your script with the following dependencies:

import os
import cobra
import pickle
import argparse
import warnings
import symengine
from random import shuffle
from multiprocessing import cpu_count
from sys import stdout
from copy import deepcopy
from subprocess import call
from cobra.util import solver
from cobra.manipulation.delete import *

Then, load your universal.pickle file using:

filename = '/universal.pickle'
with open(filename, 'rb') as f: universal = pickle.load(f)

Now you can freely add, change, or remove reactions and metabolites from the universal reaction database. You can add, remove, or modify reactions and metabolites from the universal database in the same way you would add, remove, or modify reactions and metabolites in a GENRE. Some examples of how to do this are below, but you can see further examples in the cobrapy documentation.

If you want to change the formula, of a specific metabolite formula you can do so like below

universal.metabolites.cpd17677_e.formula = 'C17H24O2'

universal.metabolites.cpd17677_e.name = 'new_name'

You can add reactions to the universal database by defining the reaction, and then adding it to the model.

universal.reaction.add_metabolites({
    malACP_c: -1.0,
    h_c: -1.0,
    ddcaACP_c: -1.0,
    co2_c: 1.0,
    ACP_c: 1.0,
    omrsACP_c: 1.0
})

universal.add_reactions([reaction])

Examples of removing metabolites and reactions from the universal database are shown below.
For a reaction named rxn:

universal.remove_reactions(rxn)

For a metabolite named met:

universal.remove_metabolites(met)

After making modifications to the universal database, you can replace the existing universal database using the following command or similar:

pickle.dump(universal, open(script_path + '/universal.pickle', 'wb'))

Thank you for your interest in Reconstructor. If you have any additional questions please email [email protected].

If you encounter any problems, please file an issue along with a detailed description.

Distributed under the terms of the MIT license, "reconstructor" is free and open source software