rna-torsionbert 0.1.3

Add your description here

RNA-TorsionBERT

RNA-TorsionBERT is a 86.9 MB parameter BERT-based language model that predicts RNA torsional and pseudo-torsional angles from the sequence.

RNA-TorsionBERT is a DNABERT model that was pre-trained on ~4200 RNA structures.

It provides improvement of MCQ over the previous state-of-the-art models like SPOT-RNA-1D or inferred angles from existing methods, on the Test Set (composed of RNA-Puzzles and CASP-RNA).

Installation

To install RNA-TorsionBERT and it's dependencies following commands can be used in terminal:

pip install -r requirements.txt 

RNA-TorsionBERT usage

To run the RNA-TorsionBERT, you can use the following command line:

python -m src.rna_torsionBERT_cli [--seq_file] [--in_fasta] [--out_path]

The arguments are the following:

• --seq_file: RNA Sequence.
• --in_fasta: Path to the input sequence fasta file.
• --out_path: Path to a .csv file where the output will be saved.

You can also import in your python code the class RNATorsionBERTCLI from src.rna_torsionBERT_cli.

TB-MCQ

TB-MCQ stands for TorsionBERT-MCQ, which is a scoring function to assess the quality of a predicted structure in torsional angle space. Given the inferred angles from the structures and the predicted angles from the model, TB-MCQ computes the quality of the predicted angles using the MCQ (mean of circular quantities) metric.

To run the TB-MCQ scoring function, you can use the following command line:

python -m src.rna_torsion_cli [--in_pdb] [--out_path]

with:

• --in_pdb: Path to the input PDB file.
• --out_path: Path to a .csv file where the output will be saved.

Docker

To run the code using Docker, you can use the following command line:

docker build -t rna_torsionbert .
docker run -it rna_torsionbert 

It will enter into a bash console where you could execute the previous commands with all the installations done.

To have example of commands, you can look at the Makefile.

Citation

@article {rna_torsionbert,
	author = {Bernard, Clement and Postic, Guillaume and Ghannay, Sahar and Tahi, Fariza},
	title = {RNA-TorsionBERT: leveraging language models for RNA 3D torsion angles prediction},
	elocation-id = {2024.06.06.597803},
	year = {2024},
	doi = {10.1101/2024.06.06.597803},
	publisher = {Cold Spring Harbor Laboratory},
	URL = {https://www.biorxiv.org/content/early/2024/07/05/2024.06.06.597803},
	journal = {bioRxiv}
}