GH29BERT

Model Introduction

This repository contains the model parameters for GH29BERT and the task-training model parameters for relating protein pre-training models. GH29BERT is a protein functional cluster prediction model designed for GH29 family sequences. It is trained based on a semi-supervised deep learning method with:

• a. 34,258 unlabeled and non-redundant GH29 sequences extracted from CAZy and Interpro databases and,
• b. 2,796 labelled sequences with 45 cluster classes based on a thorough SSN analysis.

Refer to the detailed code and data at https://github.com/ke-xing/GH29BERT.git