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.
ART package requires only a standard computer with enough RAM to support the in-memory
operations. For some problems, ART's results will improve on systems with many processor cores to
dedicate to its Markov Chain Monte Carlo (MCMC) sampling.
The ART Docker image will run on any OS that Docker supports (e.g. macOS, Linux, Windows).
Direct installs of ART have been tested on macOS and Linux. The package has been tested on the following systems:
- macOS: Mojave (10.14.1), Catalina (10.15.1)
- Linux: Debian 9 & 10
Docker is the preferred environment for running ART, as it creates reproduceable runs in a tested runtime environment. Docker also avoids the installation headaches and potential pitfalls of directly installing ART into your system Python.
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.
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 45 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.