draeger-lab / pymcadre

pymCADRE enables the reconstruction of tissue-specific metabolic models in Python using transcriptomic data and information of the network topology.

pymCADRE

Authors : Nantia Leonidou

Publication

When using pymCADRE in a research work, please cite the following work:

Leonidou, N., Renz, A., Mostolizadeh, R., & Dräger, A. (2023). New workflow predicts drug targets against SARS-CoV-2 via metabolic changes in infected cells. PLOS Computational Biology, 19(3), e1010903.

Overview.

The pymCADRE tool is an advanced re-implementation of the metabolic Context-specificity Assessed by Deterministic Reaction Evaluation (mCADRE) algorithm in Python. It constructs tissue-specific metabolic models by leveraging gene expression data and literature-based evidence, along with network topology information.

The reactions within the generic global model are being ranked, and the ones with the lowest supporting evidence for the tissue of interest are given the highest priority for removal:

GM, C, NC, P, Z, model_C = rank_reactions(model, G, U, confidence_scores, C_H_genes, method)

If the generic functionality test is passed, the model undergoes pruning, which results in a context-specific reconstruction:

PM, cRes = prune_model(GM, P, C, Z, eta, precursorMets, salvage_check, C_H_genes, method)

Installation

pip install pymcadre

Import module and sub-modules

import pymCADRE

# sub-module example
from pymCADRE.rank import *

Prerequisites

This tool has the following dependencies:

python >=3.8.5

Packages:

• pandas
• numpy
• cobra
• requests
• os

Input data

• model: COBRA model structure for the metabolic model of interest
• precursorMets: list of precursor, key, metabolites in form of .txt file
• confidence_scores: literature/experimental-based confidence assigned to reactions in model

Tissue-specific expression evidence:

• G: list of Entrez IDs for all genes in model
• U: list of ubiquity scores calculated for all genes in model

Optional Inputs

• salvageCheck: flag whether to perform a functional check for the nucleotide salvage pathway (1) or not (0)
• C_H_genes: list with Entrez IDs for genes with particularly strong evidence of activity in the tissue of interest
• method: method to use internal optimizations, (1) flux variability analysis or (2) fastcc

Outputs

• PM: pruned COBRA tissue-specific model
• GM: COBRA model after removing blocked reactions from the input global model
• C: core reactions in GM
• NC: non-core reactions in GM
• Z: reactions with zero expression across all samples after binarization
• model_C: core reactions in the generic model (including blocked reactions)
• pruneTime: total reaction pruning time
• cRes: result of model checks (consistency/function) during pruning

Usage

To run pymCADRE, execute the notebook named main_pymcadre.ipynb or the python script named pymcadre.py. The scripts can be modified to the preferred parameters and input files. Jupyter notebooks with test runs and test scripts are also provided as reference points.

Additional material

PREDICATE (Prediction of Antiviral Targets):

Steps:

• introduction of mutations in the reference sequence based on the protein sequences
• calculation of the necessary stoichiometric coefficients for the final virus biomass functions
• target detection using two approaches: reaction knock-outs and the host-derived enforcement
• visualizations that could give insights into the dataset and a better understanding of the results.

The tool can be applied to either one or more nucleotide sequences and all existing RNA viruses. This makes it particularly advantageous and time-saving when studying multiple variants of a single virus. The number of genomic input sequences equals the number of the calculated VBOF.

To run the tool, set the constant variables to the file pathways where the desired files are stored.

pymcadre's Issues

Unable to download pymCADRE with 'pip install pymcadre'

Dear Dräger Lab members,

Recently, I am planning to use pymCADRE to construct some tissue-specific GEM, But when I tried to install pymCADRE, it failed.

Could you help me to solve this issue? pymCADRE is a very good tool, I really hope that I can use it in my research.

Thanks so much

RuntimeWarning Error During Execution of pymCADRE's 'rank_reactions'

Hi! I'm currently trying to apply pymCADRE to build a context-specific metabolic model. However, when I run the 'rank_reactions' method with my gene ID list (G), ubiquity scores (U), and reaction confidence scores specified, I'm met by a RuntimeWarning error that reads: 'All-NaN axis encountered'. Below is a screenshot of the full traceback:

What do you think could be going on here? If you would like me to send the model and inputs I'm using, I would be more than happy to do so. Any clarification/feedback would be greatly appreciated, thanks!

-Rohak

Same Model Generated by Two Different Transcriptomes

Hi @NantiaL,

I hope you're doing well! I've successfully run pymCADRE on my system, generating models for the infected and un-infected conditions, respectively. However, despite using different ubiquity scores and transcriptomes for each model run, I end up with the same generic model? I've tried implementing pruning but going through each reaction iteratively would take 16-24 hours and I'm not sure if that's worth the time. Is pruning the solution here, or is it an inherent limitation of the software?

Any clarification would be greatly appreciated, thanks!

Best,
Rohak

Stable release

A stable release would be great to have.