Announcing the implementation of CAPSIF2 on ROSIE

We’d like to announce an implementation of Protein interaction of CArbohydrates Predictor and CArbohydrate Protein Site IdentiFier 2 (CAPSIF2) app on ROSIE: https://r2.graylab.jhu.edu/

PiCAP is a deep learning method that predicts whether a protein binds to carbohydrates/glycans in biological contexts. PiCAP has been used to elucidate the human, mouse, and E. coli proteomes in terms of protein-carbohydrate binding.

CAPSIF2 is a deep learning method that predicts which residues of a provided protein binds to carbohydrates. CAPSIF2 outcompetes current models for residue predictions, such as PeSTo-Carbs.

Both models are described in our recent paper: Predictions from Deep Learning Propose Substantial Protein-Carbohydrate Interplay, which can be read on BioRxiv at https://doi.org/10.1101/2025.03.07.641884

Both apps could also be run using our new Docker container, please see https://github.com/Graylab/picap?tab=readme-ov-file#running-picap-using-the-rosettacommons-docker-container for details.