PCIPG: A comprehensive framework for protein complex identification based on a probabilistic graphical model
- 1School of Mathematics, Renmin University of China, Beijing, China
- 2School of Life Science, Beijing Institute of Technology, Beijing, China
Abstract
Protein complexes are molecular machines that execute essential cellular functions, but their computational identification remains challenging. Existing protein complex identification methods largely rely on PPI network topology, functional annotations, or protein-level biochemical evidence. Although these approaches have recovered many biologically meaningful assemblies, they are often sensitive to incomplete or noisy interactomes and provide limited mechanistic insight into the residue- and interface-level determinants of complex formation. In particular, conventional PPI-based graph representations indicate whether proteins are associated, but usually ignore how protein subunits physically interact through spatially organized residues and structural interfaces. These limitations motivate the development of computational frameworks that connect residue-scale structural cues with interactome-scale organization. Here we present PCIPG, a multi-scale probabilistic graph framework that jointly models residues, proteins, interactions and complexes. PCIPG encodes residue-level physicochemical descriptors on intra-chain contact maps, screens informative residues to construct structure-aware protein representations and propagates these representations over the PPI graph to infer a protein–complex membership matrix. To couple complex membership with sparse interaction evidence, PCIPG reconstructs the network using a zero-inflated Bernoulli–Exponential likelihood, providing a principled learning signal under missing-edge and noise regimes. Across five Saccharomyces cerevisiae benchmarks, PCIPG achieved higher average F1 and Acc than the representative baseline methods included in this study, with average improvements of 11.46% and 3.64%, respectively. On the evaluated human interactomes, PCIPG achieved the highest F1 score among the compared methods on HCT116 and HEK293T, whereas its performance on HuRI was below that of AdaPPI and ClusterONE. Embedding-guided interaction completion improved PCIPG’s performance relative to its results on the corresponding original human PPI networks. Beyond complex calling, PCIPG supports core–module mining by recovering known cores and delineating coherent accessory modules within assemblies; several predictions match previously reported functional entities, including TRAPPII- and PCNA-loading-factor–related complexes. At the residue level, residues prioritized by PCIPG show increased overlap with experimentally defined protein-binding interfaces in the evaluated structures. In a computational CFTR case study, the model generated state-dependent interaction predictions that partially overlapped with experimentally profiled wild-type and ΔF508 interaction networks. Together, PCIPG bridges residue-scale structural cues with interactome-scale organization to enable interpretable and scalable protein complex identification. Code and data are available at https://github.com/hyx-1/PCIPG. Author summary: Proteins rarely work alone. In living cells, many proteins assemble into complexes that act as molecular machines to perform essential biological functions, such as DNA replication, RNA processing, protein degradation and cellular signaling. Identifying these complexes is therefore important for understanding how cells work and how diseases arise. However, current computational methods often rely mainly on protein–protein interaction networks, which can be incomplete or noisy, and usually ignore detailed structural information at the amino-acid residue level. In this study, we developed PCIPG, a graph-based computational framework that combines residue-level protein features, structural contact information and protein interaction networks to identify protein complexes. PCIPG learns how local structural properties of proteins are related to their interactions and complex memberships, and it can also infer missing interactions in incomplete networks. Across the evaluated yeast and human datasets, PCIPG generally achieved competitive or improved protein complex identification performance compared with the representative methods included in this study, although the relative performance varied across datasets. In addition, it highlighted biologically meaningful complex cores, candidate functional modules and residues likely to participate in protein-binding interfaces. Our results suggest that integrating structural information with interaction networks can provide a more interpretable way to study protein complex organization.
Related
- RiboBA: a bias-aware probabilistic framework for robust ORF identification across diverse ribosome profiling protocols
- ProteoEM: probabilistic protein abundance estimation from iterative affinity traces
- Bone morphogenetic protein 4 meets artificial intelligence: a comprehensive review of AI-driven advances in drug discovery, structural modeling, and clinical translation
- Sample preparation for direct nanopore based protein identification via tryptic peptide recognition
- Brownotate, a Comprehensive Solution to Generate Protein Sequence Databases for Any Species
BibTeX
@article{huang2026pcipg,
title = {{PCIPG: A comprehensive framework for protein complex identification based on a probabilistic graphical model}},
author = {Yixiang Huang and Lei Yang and Jiudong Wang and Xinqi Gong},
journal = {PLOS Computational Biology},
year = {2026},
month = sep,
doi = {10.1371/journal.pcbi.1014698},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014698},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - PCIPG: A comprehensive framework for protein complex identification based on a probabilistic graphical model
AU - Yixiang Huang
AU - Lei Yang
AU - Jiudong Wang
AU - Xinqi Gong
PY - 2026
DA - 2026/09/03
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014698
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014698
AB - Protein complexes are molecular machines that execute essential cellular functions, but their computational identification remains challenging. Existing protein complex identification methods largely rely on PPI network topology, functional annotations, or protein-level biochemical evidence. Although these approaches have recovered many biologically meaningful assemblies, they are often sensitive to incomplete or noisy interactomes and provide limited mechanistic insight into the residue- and interface-level determinants of complex formation. In particular, conventional PPI-based graph representations indicate whether proteins are associated, but usually ignore how protein subunits physically interact through spatially organized residues and structural interfaces. These limitations motivate the development of computational frameworks that connect residue-scale structural cues with interactome-scale organization. Here we present PCIPG, a multi-scale probabilistic graph framework that jointly models residues, proteins, interactions and complexes. PCIPG encodes residue-level physicochemical descriptors on intra-chain contact maps, screens informative residues to construct structure-aware protein representations and propagates these representations over the PPI graph to infer a protein–complex membership matrix. To couple complex membership with sparse interaction evidence, PCIPG reconstructs the network using a zero-inflated Bernoulli–Exponential likelihood, providing a principled learning signal under missing-edge and noise regimes. Across five Saccharomyces cerevisiae benchmarks, PCIPG achieved higher average F1 and Acc than the representative baseline methods included in this study, with average improvements of 11.46% and 3.64%, respectively. On the evaluated human interactomes, PCIPG achieved the highest F1 score among the compared methods on HCT116 and HEK293T, whereas its performance on HuRI was below that of AdaPPI and ClusterONE. Embedding-guided interaction completion improved PCIPG’s performance relative to its results on the corresponding original human PPI networks. Beyond complex calling, PCIPG supports core–module mining by recovering known cores and delineating coherent accessory modules within assemblies; several predictions match previously reported functional entities, including TRAPPII- and PCNA-loading-factor–related complexes. At the residue level, residues prioritized by PCIPG show increased overlap with experimentally defined protein-binding interfaces in the evaluated structures. In a computational CFTR case study, the model generated state-dependent interaction predictions that partially overlapped with experimentally profiled wild-type and ΔF508 interaction networks. Together, PCIPG bridges residue-scale structural cues with interactome-scale organization to enable interpretable and scalable protein complex identification. Code and data are available at https://github.com/hyx-1/PCIPG. Author summary: Proteins rarely work alone. In living cells, many proteins assemble into complexes that act as molecular machines to perform essential biological functions, such as DNA replication, RNA processing, protein degradation and cellular signaling. Identifying these complexes is therefore important for understanding how cells work and how diseases arise. However, current computational methods often rely mainly on protein–protein interaction networks, which can be incomplete or noisy, and usually ignore detailed structural information at the amino-acid residue level. In this study, we developed PCIPG, a graph-based computational framework that combines residue-level protein features, structural contact information and protein interaction networks to identify protein complexes. PCIPG learns how local structural properties of proteins are related to their interactions and complex memberships, and it can also infer missing interactions in incomplete networks. Across the evaluated yeast and human datasets, PCIPG generally achieved competitive or improved protein complex identification performance compared with the representative methods included in this study, although the relative performance varied across datasets. In addition, it highlighted biologically meaningful complex cores, candidate functional modules and residues likely to participate in protein-binding interfaces. Our results suggest that integrating structural information with interaction networks can provide a more interpretable way to study protein complex organization.
SN - 1553-7358
ER -