rnaglib Official Documentation

rnaglib (RNA Geometric Library) is a Python package for studying models of RNA 3D structures.

Core Features

• Quick and detailed access to all available RNA 3D structures with annotations

• Train and benchmark deep learning models for RNA 3D structure-function tasks

• RNA graph visualization

• Create fully reproducible custom datasets and tasks

Train a model

Select your task and representation to generate sample code:

Task: Chemical Modification RNA GO Inverse Folding gRNAde Ligand Identification Protein Binding Site Binding Site Benchmark Binding Site

Representation: Graph Voxel Point Cloud Sequence

Split: Default Random Sequence Structure

Copy Code

Dataset Sizes by Task

The following plot shows the number of RNAs in each available task dataset, with inverse folding tasks (large-scale) on the left and other tasks (smaller-scale) on the right:

Get started with rnaglib

• Install

• Quickstart

Tutorials

Those tutorials are meant to give you on operational overview of the library

• Working with 2.5D graphs datasets

• Using Benchmark Tasks

Data reference

Pages to read to better understand what is our data and how we build it.

• A tour of RNA 2.5D graphs

• RNA Annotation Reference

• How is the data built?

• Available Benchmark Tasks

A peek under the hood

Pages to give you an understanding of the main objects shipping with RNAglib

• Overview

• RNADataset

• RNA Transforms

• RNADataset Transforms

• Task

Package Structure

• rnaglib.dataset: custom RNA PyTorch dataset implementations

• rnaglib.tasks: prediction tasks for ML benchmarking.

• rnaglib.transforms: process and modify RNA data

• rnaglib.learning: learning routines and pre-built GCN models for the easiest use of the

• rnaglib.prepare_data: processes raw PDB structures and builds a database of 2.5D graphs with full structural annotation package.

• rnaglib.drawing: utilities for visualizing 2.5D graphs

Source Code and Contact

• Source Code.

• Contact rnaglib@cs.mcgill.ca

• X.

Associated Repositories

RNAmigos : a research paper published in Nucleic Acid Research that demonstrates the usefulness of 2.5D graphs for machine learning tasks, exemplified onto the drug discovery application.

VeRNAl : a research paper published in Bioinformatics that uses learnt vector representations of RNA subgraphs to mine structural motifs in RNA.

References

1. Leontis, N. B., & Zirbel, C. L. (2012). Nonredundant 3D Structure Datasets for RNA Knowledge Extraction and Benchmarking. In RNA 3D Structure Analysis and Prediction N. Leontis & E. Westhof (Eds.), (Vol. 27, pp. 281–298). Springer Berlin Heidelberg. doi:10.1007/978-3-642-25740-7\_13

Indices and tables

• Index

• Module Index

• Search Page