rnaglib.tasks

Task objects hold everything you need to feed a prediction model and evaluate its performance on a variety of tasks, as well as to easily implement your own tasks.

Abstract classes

Subclass these to create your own tasks.

Task(root[, debug, in_memory, recompute, ...])

Abstract class for a benchmarking task using the rnaglib datasets. This class handles the logic for building the underlying dataset which is held in an rnaglib.dataset.RNADataset object. Once the dataset is created, the splitter is invoked to create the train/val/test indices. Tasks also define an evaluate() function to yield appropriate model performance metrics.

ClassificationTask(**kwargs)

ResidueClassificationTask([additional_metadata])

RNAClassificationTask([additional_metadata])

RNA-level Classification

These tasks take as input an RNA and predict a property of the whole molecule.

RNAGo([size_thresholds])

Predict the GO terms associated with the Rfam family of a given RNA chain.

Residue-level Classification

These tasks take as input an RNA and predict a property of each residue of the molecule.

BindingSite([cutoff, size_thresholds])

Predict the RNA residues which are the most likely to be part of binding sites for small molecule ligands

BenchmarkBindingSite([cutoff])

Version of RNA-Site implemented using the data and splitting of the experiment by Su et al. (2021).

ChemicalModification([size_thresholds])

Residue-level binary classification task to predict whether a given residue is chemically modified.

ProteinBindingSite([size_thresholds])

The job is to predict a binary variable at each residue representing the probability that a residue belongs to a protein-binding interface

InverseFolding([size_thresholds])

RNA design task, taking as input the structures with the identity of the residues masked and trying to find it back

gRNAde([size_thresholds])

This class is a subclass of InverseFolding and is used to train a model on the gRNAde dataset.

Substructure-level Classification

Classification to predict properties of substructures of a whole molecule (e.g. binding pockets)

LigandIdentification([size_thresholds, ...])

Binding pocket-level task where the job is to predict the (small molecule) ligand which is the most likely to bind a binding pocket with a given structure

Substructure-level Regression

Regression to predict properties of substructures of a whole molecule (e.g. binding pockets)

VirtualScreening(root[, ligand_framework, ...])

RNA binding pocket-small molecule binding affinity prediction