Improving Enzyme Prediction with Chemical Reaction Equations by Hypergraph-Enhanced Knowledge Graph Embeddings

TL;DR

This work tackles data sparsity in enzyme–substrate prediction by leveraging abundant chemical reaction equations. It introduces Hyper-Enz, a hypergraph-enhanced KG embedding framework that represents reactions as (educts, enzyme, products) triples and models high-order compound interactions via a two-level hypergraph and a hypergraph transformer. A three-expert decision framework fuses equation-level signals, incomplete reactions, and enzyme–substrate pair information to produce robust enzyme predictions. Across equation-level and enzyme–substrate pair tasks, Hyper-Enz significantly outperforms strong baselines, demonstrating the value of structured domain knowledge and multi-source integration for practical enzyme discovery and metabolic engineering.

Abstract

Predicting enzyme-substrate interactions has long been a fundamental problem in biochemistry and metabolic engineering. While existing methods could leverage databases of expert-curated enzyme-substrate pairs for models to learn from known pair interactions, the databases are often sparse, i.e., there are only limited and incomplete examples of such pairs, and also labor-intensive to maintain. This lack of sufficient training data significantly hinders the ability of traditional enzyme prediction models to generalize to unseen interactions. In this work, we try to exploit chemical reaction equations from domain-specific databases, given their easier accessibility and denser, more abundant data. However, interactions of multiple compounds, e.g., educts and products, with the same enzymes create complex relational data patterns that traditional models cannot easily capture. To tackle that, we represent chemical reaction equations as triples of (educt, enzyme, product) within a knowledge graph, such that we can take advantage of knowledge graph embedding (KGE) to infer missing enzyme-substrate pairs for graph completion. Particularly, in order to capture intricate relationships among compounds, we propose our knowledge-enhanced hypergraph model for enzyme prediction, i.e., Hyper-Enz, which integrates a hypergraph transformer with a KGE model to learn representations of the hyper-edges that involve multiple educts and products. Also, a multi-expert paradigm is introduced to guide the learning of enzyme-substrate interactions with both the proposed model and chemical reaction equations. Experimental results show a significant improvement, with up to a 88% relative improvement in average enzyme retrieval accuracy and 30% improvement in pair-level prediction compared to traditional models, demonstrating the effectiveness of our approach.

Figures (6)

• Figure 1: Enhancing enzyme retrieval using chemical reaction equations. Direct search faces two challenges: (1) the query substrate in Reaction ① may participate in multiple biochemical processes; (2) Reaction ② lacks enzyme annotation. Collaborative information from related reactions, such as Reaction ③, can help infer missing enzyme information.
• Figure 2: Hypergraph construction. Take two toy reaction equations $q_1: c_1 + c_2 \to c_3 + c_4$ and $q_2: c_2 + c_3 \to c_4$ as an example. Hyperedges are regarded as subsets of nodes in a heterogeneous graph with node types being educts and products. There are pairwise collaborative relations between hyperedges because of sharing educts, sharing products, and cross-sharing of educts and products.
• Figure 3: Training and prediction process of Hyper-Enz. ① We construct an equation hypergraph using both $\mathcal{Q}$ and $\mathcal{Q}'$. ② During training, a hypergraph transformer-based GCN encoder is used to generate representations for $\mathcal{S}$ and $\mathcal{P}$, and a KGE-based decoder to learn enzyme-aware embeddings. The encoder and decoder are optimized on $\mathcal{Q}$. ③ In inference, we predict missing enzymes for incomplete equations in $\mathcal{Q}'$, employing a multi-expert framework that integrates outputs from a knowledge base, a pair-level ML model, and Hyper-Enz. A decision module integrates these predictions for both equation- and pair-level enzyme prediction tasks. ④ Finally, logits from the three experts are normalized and weighted to produce a ranked top $k$ enzyme list.
• Figure 4: Effect of neighbor sampling size on model performance.
• Figure 5: Hypergraph construction during inference. Suppose that the testing educt set $S=\{c_2,c_4\}$.

Theorems & Definitions (3)

• Example 1
• Example 2
• Example 3