Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UniZyme: A Unified Protein Cleavage Site Predictor Enhanced with Enzyme Active-Site Knowledge

Chenao Li, Shuo Yan, Enyan Dai

TL;DR

This work addresses the limitation of enzyme-specific cleavage-site predictors by introducing UniZyme, a unified model that generalizes across diverse proteolytic enzymes. It achieves this by a biochemically-informed enzyme encoder that incorporates energetic frustration and 3D structure to bias transformer attention, complemented by active-site knowledge through auxiliary prediction, large-scale pretraining, and active-site-aware pooling, along with substrate encoding via ESM-2. Empirical results show strong performance in both supervised and zero-shot settings, including HIV‑1 enzyme substrates, with Ablation studies confirming the contribution of energy-based features and active-site information. The approach enables robust cleavage-site predictions for novel enzymes and broad substrate spaces, with potential impact on drug design and enzyme engineering.

Abstract

Enzyme-catalyzed protein cleavage is essential for many biological functions. Accurate prediction of cleavage sites can facilitate various applications such as drug development, enzyme design, and a deeper understanding of biological mechanisms. However, most existing models are restricted to an individual enzyme, which neglects shared knowledge of enzymes and fails generalize to novel enzymes. Thus, we introduce a unified protein cleavage site predictor named UniZyme, which can generalize across diverse enzymes. To enhance the enzyme encoding for the protein cleavage site prediction, UniZyme employs a novel biochemically-informed model architecture along with active-site knowledge of proteolytic enzymes. Extensive experiments demonstrate that UniZyme achieves high accuracy in predicting cleavage sites across a range of proteolytic enzymes, including unseen enzymes. The code is available in https://anonymous.4open.science/r/UniZyme-4A67.

UniZyme: A Unified Protein Cleavage Site Predictor Enhanced with Enzyme Active-Site Knowledge

TL;DR

This work addresses the limitation of enzyme-specific cleavage-site predictors by introducing UniZyme, a unified model that generalizes across diverse proteolytic enzymes. It achieves this by a biochemically-informed enzyme encoder that incorporates energetic frustration and 3D structure to bias transformer attention, complemented by active-site knowledge through auxiliary prediction, large-scale pretraining, and active-site-aware pooling, along with substrate encoding via ESM-2. Empirical results show strong performance in both supervised and zero-shot settings, including HIV‑1 enzyme substrates, with Ablation studies confirming the contribution of energy-based features and active-site information. The approach enables robust cleavage-site predictions for novel enzymes and broad substrate spaces, with potential impact on drug design and enzyme engineering.

Abstract

Enzyme-catalyzed protein cleavage is essential for many biological functions. Accurate prediction of cleavage sites can facilitate various applications such as drug development, enzyme design, and a deeper understanding of biological mechanisms. However, most existing models are restricted to an individual enzyme, which neglects shared knowledge of enzymes and fails generalize to novel enzymes. Thus, we introduce a unified protein cleavage site predictor named UniZyme, which can generalize across diverse enzymes. To enhance the enzyme encoding for the protein cleavage site prediction, UniZyme employs a novel biochemically-informed model architecture along with active-site knowledge of proteolytic enzymes. Extensive experiments demonstrate that UniZyme achieves high accuracy in predicting cleavage sites across a range of proteolytic enzymes, including unseen enzymes. The code is available in https://anonymous.4open.science/r/UniZyme-4A67.

Paper Structure

This paper contains 33 sections, 17 equations, 8 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Preliminaries of Protein Hydrolysis
  4. Preliminaries of Active Sites in Enzyme
  5. Problem Definition
  6. Methodology
  7. Biochemically-Informed Enzyme Encoder
  8. Integrating Energy and 3D Position in Transformer.
  9. Enhancing Enzyme Representation Learning with Active Site Knowledge
  10. Cleavage Site Prediction
  11. Final Objective Function
  12. Experiments
  13. Experimental Setup
  14. Dataset Collection.
  15. Baseline Methods.
  16. ...and 18 more sections

Figures (8)

  • Figure 1: Illustration of the enzyme-catalyzed protein hydroysis.
  • Figure 2: Architecture of biochemically-informed enzyme encoder and overall framework of .
  • Figure 3: Visualizations of predicted substrate cleavage sites for HIV-1 enzymes. Predicted cleavage sites are in red color.
  • Figure 4: Ablation studies on supervised and zero-shot settings.
  • Figure 5: Hyperparameter sensitivity analysis.
  • ...and 3 more figures