This repository hosts the source code for the paper:

Designing Molecular RNA Switches with Restricted Boltzmann Machines

by Jorge Fernandez-de-Cossio-Diaz, Pierre Hardouin, Francois-Xavier Lyonnet du Moutier, Andrea Di Gioacchino, Bertrand Marchand, Yann Ponty, Bruno Sargueil, Rémi Monasson, Simona Cocco

bioRxiv preprint: https://www.biorxiv.org/content/10.1101/2023.05.10.540155v2.abstract

If you use this code, please cite this paper (or you can use the included CITATION.bib).

Setup

The code is organized as a Julia package. If you haven't setup Julia on your computer, you can install it following the simple instructions from the official website. We have tested our package on the latest Julia version at the time of writing (v1.10.4). It should work on future non-breaking versions (v1.x.y) of Julia as well.

After having installed Julia, clone this repository locally using git, on a local directory, by running the following command from the terminal:

bash git clone https://github.com/cossio/SamApp2024.jl.git

Then, navigate to this directory in your terminal, start Julia (by running julia in the terminal), and then activate the included Project.toml by running the command:

julia julia> import Pkg julia> Pkg.activate(pwd()) # activate the project in the current directory julia> Pkg.instantiate() # install all the dependencies

The last command will install all the dependencies of this package.

Next, we will configure storage for downloading data from Rfam. Create a file named LocalPreferences.toml in the root directory of this package, and add the following lines:

toml [Rfam] RFAM_DIR = "[path to local Rfam directory]" RFAM_VERSION = "14.7"

where [path to local Rfam directory] should be replaced with the path to some local directory in your system, where the package will place data downloaded from Rfam.

Lastly, import the package within a Julia session by running:

julia julia> import SamApp2024

You can use this command in the Julia REPL. The first time you import it, it might some time for precompilation (but future imports should be faster).

At this point you can call any of the functions in the package.

Running the Pluto notebooks

The included Pluto notebooks contain the code for the downstream analysis done in the paper (which depend on the SamApp2024 package). For more information about Pluto notebooks, see: https://plutojl.org.

To setup Pluto, start a fresh Julia session, and then run the following commands:

julia julia> import Pkg julia> Pkg.add("Pluto") julia> import Pluto; Pluto.run()

This starts a Pluto server. Now open your browser, and navigate to localhost:1234, which is the default address where Pluto is listening. You can then open the notebooks by navigating to the pluto directory in the repository, and opening the .jl files.

Related packages

The code implementing Restricted Boltzmann machines (training, sampling, and other functions) is provided in a separate package: https://github.com/cossio/RestrictedBoltzmannMachines.jl. Note that this package will be installed automatically as a dependency of this repository by the Julia package manager.

The code for simulations of other riboswitch families (reported in Supplementary Materials of the paper cited above) can be found at the following repository: https://github.com/cossio/RiboswitchesSimulations20240620.jl.

For an alternative implementation of Restricted Boltzmann machines in Python, see https://github.com/jertubiana/PGM. In particular, we have prepared a repository at https://github.com/cossio/SamApp2024Py, showing how to use this Python package to train an RBM on RNA sequences from the RF00162 family.

In addition, the following repository: https://github.com/2024-RNA-Chapter/2024-01-09-SAM-RBM-example, contains a self-contained example of training an RBM on an RNA family from Rfam. Two notebooks are included: one for training on CPU and another on GPU.

A Google colab notebook is also available showing how to train an RBM on the RF00162 family with a GPU.

Issues

If you encounter any problems, or need help setting up this package, please open an issue in this repository.