gRNAde: Geometric Deep Learning for RNA Design
This release is mainly centred around using gRNAde in practice. We have included a comprehensive set of checkpoints used to benchmark the model as well as for general purpose usage (trained on all of RNAsolo). Additionally, we have included functionality for partial sequence re-design constraints as well as a design notebook for visualizing and evaluating gRNAde designed RNA sequences. This can be used for deciding what set of sequences should be sent to a wet lab, for instance.
This release updates the gRNAde codebase significantly across data handling, modelling, evaluation, as well as user interface. The repository also includes pretrained model checkpoints and a detailed tutorial to enable people to use gRNAde for their own RNA designs.
Unfortunately, the arXiv paper is still not fully up-to-date with all the developments of this release. However, the code and tutorial are both extensively documented in hopes of providing sufficient information about methodology.
This release contains the state of the gRNAde codebase used for experiments in v1-v3 of the arXiv preprint.
Both the gRNAde codebase and paper are under active development; a more recent version of the code and models is available in the main branch, but the manuscript may not be updated with the latest results.
First public release of gRNAde