dnngior 0.0.5

build high-quality genome-scale metabolic model by using a deep neural network to guide gapfilling

A novel way to gapfill metabolic models

Installation instructions

To run the dnngior gapfiller, the Gurobi solver is mandatory.

pip install gurobipy

To use gurobi, you need a license. If you are an acedemic, you may get a license for free.

Once you have successfully installed gurobi, you are ready to install the dnngior gapfiller.

pip install dnngior

Optionally, you may need to also get Tensorflow (or through conda) in case you would like to use the NN_Trainer.

How to use

Gapfilling models is done using the Gapfill class:

import dnngior.gapfill_class.Gapfill  
Gapfill(path_to_model)

You may find examples of gap-filling a genome scale reconstruction (GEM) with dnngior with a complete or a defined medium in this example notebook. dnngior can gapfill both ModelSEED and BiGG models, to gapfill BiGG models you need to specify modeltype.

Gapfill(path_to_BiGG_model, modeltype='BiGG')

Custom Networks

By default dnngior uses an universally trained network capable of accurate predictions under most circumstances. If desired, it is possible to change the Neural Network you want to use during gapfilling:

Gapfill(path_to_model, trainedNNPath=path_to_NN)

You can train your own Neural Network following this tutorial: example training NN.

Alternatively you can find additional custom Neural Networks for several taxonomic groups: Custom Networks. Upon request additional specially trained networks can be made available for specific biomes or taxonomic groups.

License

Please see License

Cite

For more about the methods used in this library and in case you are using it, please cite: Boer, M.D., Melkonian, C., Zafeiropoulos, H., Haas, A.F., Garza, D.R. and Dutilh, B.E., 2024. Improving genome-scale metabolic models of incomplete genomes with deep learning. iScience, 27(12).