Selene: a PyTorch-based deep learning library for biological sequence-level data

Abstract

To enable the application of deep learning in biology, we present Selene (https://selene.flatironinstitute.org/), a PyTorch-based deep learning library for fast and easy development, training, and application of deep learning model architectures for any biological sequences. We demonstrate how Selene allows researchers to easily train a published architecture on new data, develop and evaluate a new architecture, and use a trained model to answer biological questions of interest.

Footnotes

* This manuscript has been updated with a paragraph on related work and to add links to new tutorials (e.g. training a regression model with Selene: https://github.com/FunctionLab/selene/blob/master/tutorials/regression_mpra_example/regression_mpra_example.ipynb) and documentation (https://selene.flatironinstitute.org/overview/cli.html). The library itself has also been updated to version 0.2.0 (https://github.com/FunctionLab/selene/releases/tag/0.2.0), so there are minor modifications to the text to reflect this.