Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EnzyPGM: Pocket-conditioned Generative Model for Substrate-specific Enzyme Design

Zefeng Lin, Zhihang Zhang, Weirong Zhu, Tongchang Han, Xianyong Fang, Tianfan Fu, Xiaohua Xu

TL;DR

EnzyPGM tackles substrate-specific enzyme design by jointly generating enzymes and substrate-binding pockets conditioned on functional priors and substrates. It introduces two novel components, Residue–Function Fusion (RFF) and Residue-atom Bi-scale Attention (RBA), to fuse EC priors with SE(3)-equivariant pocket refinement and to model pocket–substrate interactions at bi-scale resolutions. The authors curate EnzyPock, a large enzyme–pocket dataset, and demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock and EnzyBench, with improved binding energy and structural validity, while maintaining robust generalization to unseen EC families and substrates. This work provides a practical pathway for substrate-specific enzyme design by explicitly modeling pocket–substrate interactions and incorporating functional priors into generation.

Abstract

Designing enzymes with substrate-binding pockets is a critical challenge in protein engineering, as catalytic activity depends on the precise interaction between pockets and substrates. Currently, generative models dominate functional protein design but cannot model pocket-substrate interactions, which limits the generation of enzymes with precise catalytic environments. To address this issue, we propose EnzyPGM, a unified framework that jointly generates enzymes and substrate-binding pockets conditioned on functional priors and substrates, with a particular focus on learning accurate pocket-substrate interactions. At its core, EnzyPGM includes two main modules: a Residue-atom Bi-scale Attention (RBA) that jointly models intra-residue dependencies and fine-grained interactions between pocket residues and substrate atoms, and a Residue Function Fusion (RFF) that incorporates enzyme function priors into residue representations. Also, we curate EnzyPock, an enzyme-pocket dataset comprising 83,062 enzyme-substrate pairs across 1,036 four-level enzyme families. Extensive experiments demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock. Notably, EnzyPGM reduces the average binding energy of 0.47 kcal/mol over EnzyGen, showing its superior performance on substrate-specific enzyme design. The code and dataset will be released later.

EnzyPGM: Pocket-conditioned Generative Model for Substrate-specific Enzyme Design

TL;DR

EnzyPGM tackles substrate-specific enzyme design by jointly generating enzymes and substrate-binding pockets conditioned on functional priors and substrates. It introduces two novel components, Residue–Function Fusion (RFF) and Residue-atom Bi-scale Attention (RBA), to fuse EC priors with SE(3)-equivariant pocket refinement and to model pocket–substrate interactions at bi-scale resolutions. The authors curate EnzyPock, a large enzyme–pocket dataset, and demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock and EnzyBench, with improved binding energy and structural validity, while maintaining robust generalization to unseen EC families and substrates. This work provides a practical pathway for substrate-specific enzyme design by explicitly modeling pocket–substrate interactions and incorporating functional priors into generation.

Abstract

Designing enzymes with substrate-binding pockets is a critical challenge in protein engineering, as catalytic activity depends on the precise interaction between pockets and substrates. Currently, generative models dominate functional protein design but cannot model pocket-substrate interactions, which limits the generation of enzymes with precise catalytic environments. To address this issue, we propose EnzyPGM, a unified framework that jointly generates enzymes and substrate-binding pockets conditioned on functional priors and substrates, with a particular focus on learning accurate pocket-substrate interactions. At its core, EnzyPGM includes two main modules: a Residue-atom Bi-scale Attention (RBA) that jointly models intra-residue dependencies and fine-grained interactions between pocket residues and substrate atoms, and a Residue Function Fusion (RFF) that incorporates enzyme function priors into residue representations. Also, we curate EnzyPock, an enzyme-pocket dataset comprising 83,062 enzyme-substrate pairs across 1,036 four-level enzyme families. Extensive experiments demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock. Notably, EnzyPGM reduces the average binding energy of 0.47 kcal/mol over EnzyGen, showing its superior performance on substrate-specific enzyme design. The code and dataset will be released later.
Paper Structure (24 sections, 14 equations, 4 figures, 4 tables)

This paper contains 24 sections, 14 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. EnzyPGM: Pocket-conditioned Enzyme Generative Model
  4. Substrate-specific Enzyme Design
  5. Overall Framework
  6. Residue–Function Fusion (RFF)
  7. Residue-atom Bi-scale Attention (RBA)
  8. (I) Intra-Residue Attention.
  9. (II) Residue–Atom Attention.
  10. Feature and Coordinate Updates.
  11. Training Objective
  12. Experiments
  13. EnzyPock: Dataset Curation
  14. Experimental Settings
  15. Quantitative Results
  16. ...and 9 more sections

Figures (4)

  • Figure 1: Four cases of enzyme design. None of the enzymes in (b), (c), and (d) bind to the substrate because there are no pockets in (b), and atom clashes in (c) and (d), which are represented by red.
  • Figure 2: Overview of EnzyPGM. (a) EnzyPGM takes enzyme function-conversed sites as a masked input sequence, conditioned on EC number and substrate atom, and predicts the complete enzyme and pocket sites via RFF and RBA modules. The mask sites at the input are represented by [M] both in sequence and coordinate space, their prediction results are indicated in red at the output. (b) present the specific process of the intra-residue and residue-atom attention, the feature and coordinate updates.
  • Figure 3: Case Study. Visualization of ground-truth, EnzyPGM-generated, and 2 baseline-generated enzyme-substrate complexes in order. Each panel displays the enzyme pocket (blue) and substrate (yellow), alongside key metrics: the Vina score (binding affinity) and the number of enzyme-substrate interactions: hydrogen bond (HB), $\pi$-cation, $\pi$-$\pi$ stack, and salt bridge (Salt).
  • Figure 4: The different analysis on EnzyPGM.