Sequence-based coevolutionary prediction of species-specific interactomes
4 affiliations
- 1Computational Biology Group, Life Sciences Department, Barcelona Supercomputing Center, Barcelona, Spain
- 2Evolutionary Systems Biophysics Group, Life Sciences Department, Barcelona Supercomputing Center, Barcelona, Spain
- 3Universitat Pompeu Fabra (UPF), Barcelona, Spain
- 4European Commission, Joint Research Centre (JRC), Geel, Belgium
Abstract
Evolutionarily conserved protein-protein interactions (PPIs) reveal fundamental biological processes. However, predicting protein function solely from these interactions provides an incomplete picture of biological systems. Computational methods for predicting PPIs often struggle due to limited functional annotations in databases, making it difficult to fully understand unique biological systems. This study introduces ContextMirror2.0 (CM2.0), a coevolution-based method designed to address these limitations. Coevolutionary approaches, unlike supervised machine learning methods, do not rely on labelled datasets, proving valuable in addressing the scarcity of data for species-specific PPIs. Around 40% of CM2.0’s top-1000 predicted PPIs for E. coli, S. enterica, and S. aureus are supported by experimental PPI databases. However, comparative analysis of the resulting functional communities further revealed both highly conserved interaction patterns and species-specific interactions. CM2.0 leverages coevolutionary information at the phylogenetic level to explore protein interaction dynamics and understand the functional consequences of species-specific variations. This information can inform further studies on the emergence of novel functions, adaptation to specific environments, and the development of targeted therapies. Author summary: Predicting how proteins interact is essential for understanding how cells work, yet most protein–protein interactions remain unknown, particularly in bacteria and other less-studied organisms. ContextMirror 2.0 is an improved version of our unsupervised computational method that detects functional relationships between proteins by comparing how their evolutionary histories change across thousands of species. Unlike machine-learning and structure-based predictors, ContextMirror 2.0 does not require training data or detailed structural information, making it suitable for analyzing complete proteomes. We applied the method to three bacterial species and found that it accurately recovers many known interactions and identifies new, species-specific protein communities that suggest potential biological functions. We also tested its performance using several benchmarking strategies, including both balanced and realistically imbalanced datasets, and showed that it consistently enriches true interactions above random expectation. Because it is scalable and interpretable, ContextMirror 2.0 can help researchers explore poorly characterized proteomes and guide future experimental studies of protein function and interaction.
Related
- Programmable sequence-resolved in situ profiling of glycoRNA-protein interactomes by GlycoRNA-Map
- Effective Resistance and Graph Neural Network Reliability in Tissue-Specific Interactomes
- Resolving context-specific protein-protein interactomes forbiological discovery and therapeutic target prioritisation
- SedaDNA metabarcoding reveals hidden histories of within-species sequence variation
- Complete plastid genome sequence of the Korean endemic species Zabelia densipila M.P. Hong, Y.C. Kim & B.Y. Lee (Caprifoliaceae) and its phylogenetic position
BibTeX
@article{fernandezmartin2026sequencebased,
title = {{Sequence-based coevolutionary prediction of species-specific interactomes}},
author = {Miguel Fernandez-Martin and Camila Pontes and Victoria Ruiz-Serra and Alfonso Valencia},
journal = {PLOS Computational Biology},
year = {2026},
month = nov,
doi = {10.1371/journal.pcbi.1014839},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014839},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Sequence-based coevolutionary prediction of species-specific interactomes
AU - Miguel Fernandez-Martin
AU - Camila Pontes
AU - Victoria Ruiz-Serra
AU - Alfonso Valencia
PY - 2026
DA - 2026/11/06
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014839
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014839
AB - Evolutionarily conserved protein-protein interactions (PPIs) reveal fundamental biological processes. However, predicting protein function solely from these interactions provides an incomplete picture of biological systems. Computational methods for predicting PPIs often struggle due to limited functional annotations in databases, making it difficult to fully understand unique biological systems. This study introduces ContextMirror2.0 (CM2.0), a coevolution-based method designed to address these limitations. Coevolutionary approaches, unlike supervised machine learning methods, do not rely on labelled datasets, proving valuable in addressing the scarcity of data for species-specific PPIs. Around 40% of CM2.0’s top-1000 predicted PPIs for E. coli, S. enterica, and S. aureus are supported by experimental PPI databases. However, comparative analysis of the resulting functional communities further revealed both highly conserved interaction patterns and species-specific interactions. CM2.0 leverages coevolutionary information at the phylogenetic level to explore protein interaction dynamics and understand the functional consequences of species-specific variations. This information can inform further studies on the emergence of novel functions, adaptation to specific environments, and the development of targeted therapies. Author summary: Predicting how proteins interact is essential for understanding how cells work, yet most protein–protein interactions remain unknown, particularly in bacteria and other less-studied organisms. ContextMirror 2.0 is an improved version of our unsupervised computational method that detects functional relationships between proteins by comparing how their evolutionary histories change across thousands of species. Unlike machine-learning and structure-based predictors, ContextMirror 2.0 does not require training data or detailed structural information, making it suitable for analyzing complete proteomes. We applied the method to three bacterial species and found that it accurately recovers many known interactions and identifies new, species-specific protein communities that suggest potential biological functions. We also tested its performance using several benchmarking strategies, including both balanced and realistically imbalanced datasets, and showed that it consistently enriches true interactions above random expectation. Because it is scalable and interpretable, ContextMirror 2.0 can help researchers explore poorly characterized proteomes and guide future experimental studies of protein function and interaction.
SN - 1553-7358
ER -