Comprehensive interaction profiling and machine learning prediction of bacteriophage infectivity across clinically diverse Pseudomonas aeruginosa
6 affiliations
- 1Environmental Genomics and Systems Biology, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
- 2Molecular and Cell Biology Department, University of California Berkeley, Berkeley, CA, USA
- 3Plant and Microbial Biology Department, University of California Berkeley, Berkeley, CA, USA
- 4Department of Bioengineering, University of California Berkeley, Berkeley, CA, USA
- 5Biological Systems and Engineering, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
- 6DOE Joint Genome Institute, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
Abstract
The rise of antibiotic-resistant bacterial infections has driven renewed interest in bacteriophage therapy, where viruses that specifically kill bacteria are used as targeted antimicrobials. Pseudomonas aeruginosa , a WHO critical-priority pathogen that causes severe infections in hospitalized and immunocompromised patients, presents a major challenge for phage therapy because of its extraordinary genetic diversity. Phages effective against one bacterial strain often fail against others, and existing cross-resistance-profiling approaches require iterative empirical testing of each new patient isolate. To establish a genome-based framework for rapid phage-isolate matching, we assembled a collection of 95 genomically diverse P. aeruginosa phages representing 20 genera and tested each against 99 genetically diverse clinical isolates, generating 9,405 infection outcome measurements. Bacterial O-antigen serotype emerged as the dominant determinant of strain susceptibility, while defense systems, anti-defense systems, and prophage burden contributed smaller strain-specific effects. The full curated multivariate model explained 47% of strain-susceptibility variance. Machine-learning models integrating these features and pangenome-derived gene clusters reached a per-strain AUROC of 0.86. In an in vivo proof-of-concept test against a single held-out strain, the ML-designed cocktail produced a ∼12-fold greater median CFU reduction than the expert-designed cocktail (q = 0.045), with both cocktails substantially reducing burden relative to the untreated control (∼113-fold for ML, ∼9-fold for CG; both q < 10□³). SHAP analysis of the model identified bacterial surface-architecture genes (LPS biosynthesis, outer membrane proteins, type IV pili) as the dominant predictors, with defense-system content modulating which specific phages succeed against a strain rather than uniformly damping susceptibility. Together, these results establish a genome-based framework for predicting phage susceptibility in genetically diverse clinical isolates.
Related
- Bacteriophage PAK_P3 genome structuration and dynamics during infection of Pseudomonas aeruginosa reveal specific interaction patterns
- Refractory Pseudomonas aeruginosa Bronchopulmonary Infection After Lung Transplantation for Common Variable Immunodeficiency Despite Maximal Treatment Including IgM/IgA-Enriched Immunoglobulins and Bacteriophage Therapy
- The bacteriophage transcription factor Gp53 suppresses cell division and regulates Pseudomonas aeruginosa physiology
- A dual-action bacteriophage protein Ped1 degrades DNA and blocks replication in Pseudomonas aeruginosa
- Bacteriophage and Antibiotic Resistance Are Positively Associated across a Phylogenetically Diverse Set of Clinical Pseudomonas aeruginosa Isolates
BibTeX
@misc{piya2026comprehensive,
title = {{Comprehensive interaction profiling and machine learning prediction of bacteriophage infectivity across clinically diverse Pseudomonas aeruginosa}},
author = {Denish Piya and Avery J. C. Noonan and Hemaa Selvakumar and Mohamad Alayouni and Sarshad Koderi Valappil and Flavien Maucourt and Isabella Murray and Madeline Svab and Collis Bousliman and Matthew Heidenblut and Britney Orihuela and Alexey Kazakov and Hans K. Carlson and Yiyan Yao and Emerson Smith and Simon Roux and Adam M. Deutschbauer and Jamie L. Inman and Adam P. Arkin and Vivek K. Mutalik},
howpublished = {bioRxiv (Cold Spring Harbor Laboratory)},
year = {2026},
month = may,
doi = {10.64898/2026.05.19.726084},
url = {https://doi.org/10.64898/2026.05.19.726084},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Comprehensive interaction profiling and machine learning prediction of bacteriophage infectivity across clinically diverse Pseudomonas aeruginosa
AU - Denish Piya
AU - Avery J. C. Noonan
AU - Hemaa Selvakumar
AU - Mohamad Alayouni
AU - Sarshad Koderi Valappil
AU - Flavien Maucourt
AU - Isabella Murray
AU - Madeline Svab
AU - Collis Bousliman
AU - Matthew Heidenblut
AU - Britney Orihuela
AU - Alexey Kazakov
AU - Hans K. Carlson
AU - Yiyan Yao
AU - Emerson Smith
AU - Simon Roux
AU - Adam M. Deutschbauer
AU - Jamie L. Inman
AU - Adam P. Arkin
AU - Vivek K. Mutalik
PY - 2026
DA - 2026/05/20
PB - bioRxiv (Cold Spring Harbor Laboratory)
DO - 10.64898/2026.05.19.726084
UR - https://doi.org/10.64898/2026.05.19.726084
AB - The rise of antibiotic-resistant bacterial infections has driven renewed interest in bacteriophage therapy, where viruses that specifically kill bacteria are used as targeted antimicrobials. Pseudomonas aeruginosa , a WHO critical-priority pathogen that causes severe infections in hospitalized and immunocompromised patients, presents a major challenge for phage therapy because of its extraordinary genetic diversity. Phages effective against one bacterial strain often fail against others, and existing cross-resistance-profiling approaches require iterative empirical testing of each new patient isolate. To establish a genome-based framework for rapid phage-isolate matching, we assembled a collection of 95 genomically diverse P. aeruginosa phages representing 20 genera and tested each against 99 genetically diverse clinical isolates, generating 9,405 infection outcome measurements. Bacterial O-antigen serotype emerged as the dominant determinant of strain susceptibility, while defense systems, anti-defense systems, and prophage burden contributed smaller strain-specific effects. The full curated multivariate model explained 47% of strain-susceptibility variance. Machine-learning models integrating these features and pangenome-derived gene clusters reached a per-strain AUROC of 0.86. In an in vivo proof-of-concept test against a single held-out strain, the ML-designed cocktail produced a ∼12-fold greater median CFU reduction than the expert-designed cocktail (q = 0.045), with both cocktails substantially reducing burden relative to the untreated control (∼113-fold for ML, ∼9-fold for CG; both q < 10□³). SHAP analysis of the model identified bacterial surface-architecture genes (LPS biosynthesis, outer membrane proteins, type IV pili) as the dominant predictors, with defense-system content modulating which specific phages succeed against a strain rather than uniformly damping susceptibility. Together, these results establish a genome-based framework for predicting phage susceptibility in genetically diverse clinical isolates.
ER -