propyly / robochem_v1.0 Goto Github PK

This repository contains the codebase and resources necessary to operate RoboChem, a self-optimizing robotic platform dedicated to chemistry optimization.

For the paper associated with the code please see here: https://chemrxiv.org/engage/chemrxiv/article-details/64809b97e64f843f41767eac

DOI: 10.26434/chemrxiv-2023-r0drq

Robochem is split over a 64-bit (GUI and BO) and 32-bit (Hardware/Platform).

Please Read below for how to get up and running.

This section consists of individual folders dedicated to managing the physical components essential for our system's operation. Each piece equipment, including the photoreactor (eagle reactor), liquid handler, mass flow controller, NMR (spinsolve NMR), phase sensors, syringe pumps, switch valves, and, has its separate folder. This modular structure allows for easy integration and modification of specific components, enabling a flexible and scalable build process. The individual components are then combined in in further individual programmes

The second section is subdivided into two essential parts:

This part encompasses the machine learning algorithms implementing Bayesian optimization using the Dragonfly package. These algorithms work tirelessly behind the scenes, optimizing chemical processes for efficiency and accuracy.

Alongside the machine learning process, a user-friendly GUI built on Streamlit orchestrates the entire system. This intuitive interface harmonizes with the optimization algorithms, empowering users to interact effortlessly with the platform. Together, these components form an accessible, self-optimizing robotic platform that streamlines the chemistry optimization process.

Two conda environments are required to run this code, one for the optimization (robochem_gui) and one for controlling the platform (robochem_platform_32bit). To create these environments please use the environment yml files:

• ML_GUI\\gui_environment.yml
• Platform_\\platform_environment.yml Please don't change the names of the environments as they are currently used in the code.

To do this, run the following commands from the root directory in the terminal:

cd ML_GUI
conda env create -f gui_environment.yml
cd ..
cd Platform_
conda env create -f platform_environment.yml

For more information on setting up conda see: https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html

If the campaign shall be run from the GUI, first the GUI has to be activated. When all information is filled in (and the Liquid Handler is filled with the stocksolutions according to the by the GUI generated excel-file) the platform can be run by pressing the run button on the Run Platform page. This will call the run_optimization_from_gui.py file which will call the Dragonfly_BO.py file which in turn will use os.popen to run the run_platform.py file in the Platform_ folder.

Change directory and then active the conda environment robochem_gui, and then run the GUI home page:

cd ML_GUI -> conda activate robochem_gui -> streamlit run ??_Home.py

When streamlit is run, the main page is 🏡_Home.py. Every additional page is found in the folder Pages. The pages are order alphabetically. When the platform is run, the result shall be updated and plotted after each run. The figures are generated in Visualization.py and the png-files figure_hypervolyme.png and figure_objectives.png are updated and these are the files the streamlit app is continuously calling. Before the first experiment has been finished the figures are replaced with the figure Campaign has started.png instead.

If one wants to run the platform directly from the code, one can write the space directly into the file run_optimization.py. An example of how that could look like is found in this file. The other files such as the excel-file for the Liquid Handler then has to manually be generated and put in the correct place. When run it will call the Dragonfly_BO.py file which in turn will use subprocess.run to run the run_platform.py file in the Platform_ folder.