rna-synthub 0.1.0

Add your description here

RNA-SyntHub 🧬

Meta-scoring pipeline to prepare curated synthetic RNA structure data

RNA-SyntHub is a pipeline applied to create curated RNA structures data. It has been applied to a set of 447,402 synthetic RNA structures and enforced by predictions from RNAComposer and Boltz-1.

Repo structure:

To download the data of RNA-SyntHub (structures from RFDiffusion, RNAComposer and Boltz-1), you can find the data on this link. The repo is structured as follows:

• data/: folder containing the data used in the repo (figures, example pdb files, etc.) as well as the .pdb structures
  • img/: folder containing the figures used in the README
  • pdb/: folder containing example .pdb files
  • examples: folder containing example output files of the different steps of the pipeline
  • metadata: folder containing example extracted information from the RNA-SyntHub dataset
  • times: folder containing times benchmark on the subset of 1000 structures
• src/: folder containing the source code of the pipeline and the visualisation
• Makefile: file to run the different steps of the pipeline and visualisation

Installation

To compute molprobity scores, you need to install RNAqua using:

make install_rnaqua

To compute alignment, you need to install mlocarna and cd-hit. You can go to src/rna_synthub/locarna_helper.py to change the path of the binaries if needed. For cd-hit, you need to compile it first using:

make install_cd_hit

You can install the visualisation using pip:

pip install rna_synthub

Usage

Pipeline steps

If you want to reproduce the pipeline, please go the the src/analyze/script_example.py where there is a step-by-step example of the pipeline usage. You just need to uncomment the main function and run the script you want. Examples are provided with 10 .pdb files. The different steps of the pipeline are as follows:

def main():
    ScriptExample.compute_n_meta()
    ScriptExample.compute_filter()
    ScriptExample.compute_2d()
    ScriptExample.compute_clustering()

You can run the script using:

make run_example

or, if you installed the library using pip:

python synt_example

Viz steps

To run the code to obtain the different visualisation, you can use:

make run_viz

or, if you installed the library using pip:

python viz_cli

You can also have a look at the src/viz/viz_cli.py to choose which visualisation you want to run.

def main():
    VizCLI.viz_funnel()
    VizCLI.viz_filter()
    VizCLI.viz_best_worst()
    VizCLI.viz_distance()
    VizCLI.viz_sf()
    VizCLI.viz_times()
    VizCLI.viz_angles()

Citation

Authors

• Clement Bernard
• Marek Justyna
• Guillaume Postic
• Maciej Antczak
• Fariza Tahi
• Marta Szachniuk