Hierarchical Contrastive Learning for Protein-Protein Interaction Prediction Across Organisms
Shiyi Liu1,2, Buwen Liang3, Yuetong Fang1
1Function Hub, The Hong Kong University of Science and Technology (Guangzhou), Guangzhou 511453, China.
International Journal of Molecular Sciences
|July 28, 2026
Summary
HIPPO, a novel hierarchical contrastive learning framework, enhances protein-protein interaction (PPI) prediction by integrating structured biological knowledge. This approach improves accuracy across diverse datasets, outperforming existing methods.
Area of Science:
- Bioinformatics
- Computational Biology
- Systems Biology
Background:
- Biological data is rapidly growing in scale and complexity.
- Existing protein-protein interaction (PPI) prediction methods often fail to model hierarchical biological relationships.
- Contrastive learning offers a framework for integrating heterogeneous biological information.
Purpose of the Study:
- Introduce HIPPO (Hierarchical Protein-Protein interaction prediction across Organisms), a novel hierarchical contrastive learning framework for PPI prediction.
- To align protein sequence representations with structured biological attributes for improved prediction accuracy.
- To evaluate HIPPO's performance on intra-species and host-pathogen interaction benchmarks.
Main Methods:
- Developed HIPPO, a hierarchical contrastive learning framework.
- Integrated structured biological attributes (families, clans, functional annotations) into representation learning.
- Evaluated HIPPO on intra-species PPI datasets, host-pathogen interaction benchmarks, and using leave-one-virus-family-out evaluation.
- Conducted ablation experiments and attention-based residue attribution analysis.
Main Results:
- HIPPO improved average micro-F1 by 2.9% on intra-species benchmark PPI datasets compared to the best baseline.
- Achieved the highest AUROC (0.731) and second-best AUPRC (0.332) on the host-pathogen interaction benchmark.
- Obtained the best AUROC on Papillomaviridae (0.603) and Retroviridae (0.612) in leave-one-virus-family-out evaluation, demonstrating family-dependent transfer behavior.
- Ablation studies confirmed the contribution of hierarchical feature integration.
Conclusions:
- Structured biological knowledge significantly improves representation learning for PPI prediction.
- HIPPO demonstrates superior performance across diverse and imbalanced biological datasets.
- The framework's ability to model hierarchical relationships enhances prediction accuracy and interpretability.
Related Concept Videos
02:04Protein-protein Interfaces
Many proteins form complexes to carry out their functions, making protein-protein interactions (PPIs) essential for an organism's survival. Most PPIs are stabilized by numerous weak noncovalent chemical forces. The physical shape of the interfaces determines the way two proteins interact. Many globular proteins have closely-matching shapes on their surfaces, which form a large number of weak bonds. Additionally, many PPIs occur between two helices or between a surface cleft and a polypeptide...
02:04Protein-Protein Interfaces
Many proteins form complexes to carry out their functions, making protein-protein interactions (PPIs) essential for an organism's survival. Most PPIs are stabilized by numerous weak noncovalent chemical forces. The physical shape of the interfaces determines the way two proteins interact. Many globular proteins have closely-matching shapes on their surfaces, which form a large number of weak bonds. Additionally, many PPIs occur between two helices or between a surface cleft and a polypeptide...
02:26Protein Networks
An organism can have thousands of different proteins, and these proteins must cooperate to ensure the health of an organism. Proteins bind to other proteins and form complexes to carry out their functions. Many proteins interact with multiple other proteins creating a complex network of protein interactions.
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
02:26Protein Networks
An organism can have thousands of different proteins, and these proteins must cooperate to ensure the health of an organism. Proteins bind to other proteins and form complexes to carry out their functions. Many proteins interact with multiple other proteins creating a complex network of protein interactions.
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
02:26Conservation of Protein Domains Over Different Proteins
Protein domains are small structurally independent units that are part of a single amino acid chain. Although these domains are often structurally independent, they may rely on synergistic effects to perform their functions as part of a larger protein. Protein domains may be conserved within the same organism, as well as across different organisms.
A limited set of protein domains often duplicate and recombine during evolution. These domains can be organized in different combinations to form...
A limited set of protein domains often duplicate and recombine during evolution. These domains can be organized in different combinations to form...
01:57Protein Complexes with Interchangeable Parts
Groups of proteins may form a complex where each protein in this complex has a different role in the overall execution of the complex’s function. Often some of the proteins in the complex can be replaced by a closely related variant to give a complex that contains many of the same components yet is functionally distinct.
The SCF ubiquitin ligase is a protein complex of five individual proteins. This complex attaches ubiquitin to other target proteins to mark them for degradation. In order to...
The SCF ubiquitin ligase is a protein complex of five individual proteins. This complex attaches ubiquitin to other target proteins to mark them for degradation. In order to...