Skip to content
/ ART Public

A machine learning tool to improve the effectiveness of strain engineering in synthetic biology

Notifications You must be signed in to change notification settings

JBEI/ART

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

73 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Automated Recommendation Tool (ART)

ART is a tool that leverages machine learning and probabilistic modeling techniques to guide metabolic engineering in a systematic fashion, without the need for a full mechanistic understanding of the biological system. Using sampling-based optimization, ART provides a set of recommended strains to be built in the next engineering cycle in order to achieve the given objective, alongside probabilistic predictions of their production levels.

Please note that this repository does not contain ART source code. For information on how to access the library see the License section.

Please find more details about ART at the ART website.

System Requirements

Hardware requirements

The ART package requires only a standard computer with enough RAM to support the in-memory operations. While slimmer installs may be possible, at the time of writing, ART developers typically dedicate 8GB RAM and 80GB of disk to the Docker Desktop instance used to develop and test the code on MacOS.

ART requires a minimum of 2 physical processor cores to run Markov Chain Monte Carlo (MCMC) sampling. For more complex problems, ART's results will improve on systems with many processor cores to dedicate to MCMC sampling (see the max_mcmc_cores parameter).

Software requirements

OS Requirements

The ART Docker image will run on any OS that Docker supports (e.g. macOS, Linux, Windows).

Direct install of ART into your system Python is not the recommended or supported workflow. Please use the included Docker image for reproducible builds, and as a reference for any direct installs you attempt on your own. While there are valid reasons to use ART externally to the included Docker image, unfortunately the ART team is not resourced to support non-Docker workflows.

In the past, direct installs of ART had been tested on macOS and Linux:

  • macOS: Mojave (10.14.1), Catalina (10.15.1)
  • Linux: Debian 9 & 10

Docker

Docker is the preferred / supported environment for running ART. Docker creates reproducible runs in a tested runtime environment. Docker also avoids the installation headaches and potential pitfalls of directly installing ART into your system Python.

Examples

An example is provided in the Limonene_Example.html file. Generating this output in a jupyter notebook should take ~5 mins on a MacBook Pro, CPU: 3.5GHz Intel Core i7, RAM: 16GB (2133MHz LPDDR3).

Additional tutorials, including real and simulated data sets, are provided in the notebooks directory.

Reference

Radivojević T., Costello Z., Workman K., Garcia Martin H., A machine learning Automated Recommendation Tool for synthetic biology, Nat Commun 11, 4879 (2020).

License

ART code is distributed under the license specified in the Noncomercial_Academic_LA.pdf file and is Patent Pending.

This license allows for free non-commercial use for academic institutions. Modification should be fed back to the original repository to benefit all users. If interested in an academic license of this type, please email [email protected] using the email address from your academic institution, and provide your github handle. You will then be added to the private github repository containing the ART source code.

A separate commercial use license is available from Berkeley Lab @ [email protected]. The license terms (10 years) are $10,000 for small businesses (less than 250 employees) and $25,000 for large businesses (more than 250 employees). Once the license is signed, interested parties will receive the information for accessing the private github repository containing the ART source code.

An evaluation license for commercial users can be obtained for 90 days of testing by filling the Evaluation_LA.pdf file and sending back to Jean Haemmerle, LBNL Licensing Associate @ [email protected]. Once the license is signed, interested parties will receive the information for accessing the private github repository containing the ART source code.

About

A machine learning tool to improve the effectiveness of strain engineering in synthetic biology

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •