The research, led by Avery J. C. Noonan and colleagues at Lawrence Berkeley National Laboratory and collaborating institutions, was built on an unusually rigorous foundation. The team systematically optimized their workflow over more than 13.2 million individual training runs, evaluating over 200 parameter settings across six published phage–host interaction datasets. Together these datasets encompass 115,037 experimentally determined interactions spanning 949 bacterial strains and 518 phages, drawn from Escherichia coli, Klebsiella, Pseudomonas and Vibrionaceae hosts. The proportion of infectious interactions within each dataset ranged from as little as 2.1 percent to 36.2 percent, a severe imbalance that mirrors the reality of phage biology, where most phages fail to infect most strains they encounter.

The core technical challenge the researchers faced is one familiar to anyone working at the intersection of genomics and machine learning: dimensionality. Each bacterial or phage genome can contribute thousands of numerical features—protein families, gene fragments or sequence k-mers—while the phenotypic data describing which phage infects which strain is sparse, binary and heavily skewed toward negative outcomes. This combination invites overfitting, in which a model memorizes quirks of its training data rather than learning generalizable rules. Many earlier tools sidestepped the problem by restricting features to genes already known to mediate infection, such as phage tail fibres or host capsule-synthesis loci. That strategy works within a single, well-studied genus but cannot transfer to emerging pathogens or poorly characterized phages, and it forecloses the discovery of novel infection determinants.

The new framework, by contrast, is deliberately phylogeny-agnostic. Genomes are represented as binary presence–absence profiles of protein families, clustered with MMSeqs2 at optimized similarity thresholds. Informative features are then selected through recursive feature elimination, and predictions are generated by an ensemble of gradient-boosted decision trees using the CatBoost algorithm, with class weights computed separately for each phage to compensate for imbalance. The team compared eight machine-learning algorithms, six feature-selection methods and multiple genome-representation schemes, finding that algorithmic choices mattered more than how genomes were encoded. Notably, the simplest representation—short 3-amino-acid k-mers—performed worst, while protein families, longer k-mers and hybrid schemes captured sufficient biological signal.

Performance was assessed through nested cross-validation under three increasingly demanding scenarios: predicting infections of unseen bacterial strains by known phages, infections of known strains by unseen phages, and interactions between entirely novel phage–strain pairs. Across datasets, the area under the receiver operating characteristic curve ranged from 0.67 to 0.94, with Matthews correlation coefficients between 0.13 and 0.54. For E. coli, the model achieved an AUROC of 0.87, essentially matching the 0.86 reported by a previous species-specific method that required phage-specific models and host features based on known E. coli infection mediators. The new approach reached comparable accuracy without any of those constraints. Performance correlated strongly with dataset size, and the weakest results came from the smallest, most imbalanced dataset—a Klebsiella matrix of only 3,658 interactions with 4.9 percent positives—suggesting the limits reflect data availability rather than flaws in the method.

The framework also translated its probability scores into practical phage-cocktail design. Phages were clustered into ‘activity groups’ using HDBSCAN density-based clustering, and the top-ranked phage from each cluster was selected, balancing predicted efficacy against mechanistic diversity. In silico tests across twenty-fold cross-validation showed that model-guided selection outperformed the common heuristic of choosing the most promiscuous phages. For single-phage selection, the models identified active phages for 66.9 percent of E. coli strains, 66.7 percent of Klebsiella strains and 39.4 percent of Vibrionaceae strains—improvements of 1.1-fold, 3.1-fold and 2.7-fold over promiscuity-based choices. Five-phage cocktails achieved bacterial coverage of up to 97.5 percent for E. coli, 87.8 percent for Klebsiella and 57.5 percent for Vibrionaceae, consistently beating baseline strategies.

Crucially, the team did not stop at computational validation. Because the phages in their training data were not available for retesting, they experimentally challenged the model with 52 previously unseen phages from the BASEL collection, spotted against 25 E. coli strains from the ECOR reference collection. Of the 1,300 tested interactions, 1,240 yielded clear phenotypes, and the model’s predictions against this genuinely novel matrix achieved an AUROC of 0.84—only marginally below its cross-validation performance of 0.87. This demonstrated that the framework generalizes to phages it had never encountered and to interaction data generated in a different laboratory with different protocols, a stringent test that many published prediction tools have never undergone.

To verify that the model’s predictions rest on real biology rather than statistical artifacts, the researchers turned to genome-wide random barcode transposon-site sequencing, or RB-TnSeq. They screened a pooled library of 3,804 gene knockouts in E. coli ECOR27 against 19 phages, identifying 51 genes whose disruption altered susceptibility. When these genetically validated infection mediators were compared with the model’s top predictive features, ranked by SHAP values, 68.6 percent showed a connection: 12 mapped directly to predictive features, 10 lay within a three-gene window of one, and 13 were functionally linked through the STRING protein-association database. Direct matches were enriched 5.0-fold above random expectation. Interestingly, the model often picked out a single regulatory gene within a larger biosynthetic pathway—porin regulators such as tsx, fadL and nfrA, or individual genes in lipopolysaccharide and capsule clusters—suggesting it captures pathway-level regulation rather than merely listing every component.

Feature interpretation also surfaced both familiar and unexpected biology. Among the top 25 predictive features in the E. coli dataset were elements tied to cell wall biosynthesis, mobile genetic elements, restriction–modification systems and genes of viral origin, alongside eight features with no clear mechanistic link that may represent uncharacterized infection mediators. Network and synteny analyses pointed to candidate roles for putrescine catabolism and regulation of the known phage receptor fepA, hypotheses the authors flag for future experimental work. Defence systems generally carried negative impacts on infection probability, as expected, while anti-defence features trended positive.

The study is candid about its limitations. Cross-genus prediction remains poor, with leave-one-genus-out AUROC values hovering barely above random, indicating that interaction determinants are largely genus-specific or that protein-family features cannot capture conserved mechanisms across deep evolutionary distances. RB-TnSeq validation is restricted to non-essential genes with strong fitness effects, and the authors suggest CRISPR interference screens as a complement for essential genes. Performance also depends on large datasets of confirmed interactions, which are slow and expensive to produce. Nevertheless, the successful merging of Klebsiella datasets from different research groups points toward community-driven, standardized data generation as a viable path forward. As antimicrobial resistance escalates, the authors argue, this framework—released as the open-source GenoPHI package—offers a scalable computational platform for rational phage-therapy design and precision microbiome engineering across clinical, agricultural and industrial settings.

Subject of Research: Machine-learning prediction of strain-level bacteriophage–host interactions from bacterial and phage genome sequences

Article Title: Phylogeny-agnostic strain-level prediction of phage–host interactions from genomes using machine learning

Article References: Noonan, A. J. C., Moriniere, L., Rivera-López, E. O., Patel, K., Pena, M., Svab, M., Kazakov, A., Deutschbauer, A., Dudley, E. G., Mutalik, V. K., & Arkin, A. P. (2026). Phylogeny-agnostic strain-level prediction of phage–host interactions from genomes using machine learning. Nature Microbiology. https://doi.org/10.1038/s41564-026-02482-5

Image Credits: AI Generated

DOI: 10.1038/s41564-026-02482-5

Keywords: bacteriophages, phage therapy, machine learning, phage–host interactions, genomics, antimicrobial resistance, strain-level prediction, RB-TnSeq, phage cocktails, microbiome engineering, CatBoost, protein families

The Pseudomonas syringae complex is a phylogenetic group comprising at least fifteen bacterial species and more than sixty host-specialized pathovars that collectively cause disease in over 180 plant species, including virtually all economically relevant crops. In highbush blueberry (Vaccinium corymbosum), infections manifest as blight of leafy tissues or cankers of woody tissues, both of which reduce plant health and productivity. Disease development depends on the transition of the bacterium from an epiphytic phase, in which cells remain adhered to plant surfaces, to an endophytic phase involving colonization of internal structures and the release of tissue-destroying virulence factors. Environmental conditions common in British Columbia—cool temperatures, high humidity and frequent precipitation—favour the growth and motility of epiphytic populations and thereby increase the likelihood of endophytic infection, making strategies that reduce surface populations before infection establishes an attractive line of defence.

The economic stakes are considerable. Global blueberry production reached roughly 2.14 billion kilograms in 2024, and Canada is the fourth-largest producer worldwide, accounting for nearly one-sixth of global output. Canadian blueberry exports were valued at approximately CAD $695 million in 2024, representing more than 65 percent of all Canadian fruit exports. Nearly 94 percent of the country’s highbush blueberry production occurs in British Columbia, where conditions favourable to the crop simultaneously favour bacterial growth and dispersal on plant surfaces. Traditional management has relied on copper-based biocides, but this approach damages the natural environment, can stunt plant growth through interference with the rhizosphere microbiome, and has been losing efficacy due to the spread of plasmid-borne copper resistance genes. Mitigation is often limited to the costly and unsustainable practice of pruning or culling infected plants.

Against this backdrop, the research team isolated bacteriophages from wastewater samples and from Pseudomonas syringae complex-infected blueberry plants collected at sites in the lower mainland and Okanagan valley regions, ultimately obtaining nineteen relatively stable phage isolates. Because phage predation by individual phage species is generally restricted to a subset of target strains and readily selects for resistant subpopulations, cocktail formulations of multiple distinct phages are typically employed to expand host range, improve efficacy and forestall resistance. The researchers used a semiquantitative spotting analysis to score infection interactions between each phage and a panel of twenty bacterial strains on a scale from zero to four, then excluded seven isolates—deemed weak specialists—that fell below median thresholds for both the number and strength of interactions. Twelve candidate phages advanced to genomic and functional characterization.

Whole-genome sequencing revealed extensive phylogenomic diversity among the candidates. Three isolates proved to be variants of a species within the genus Unosvirus, three were clonal duplicates belonging to Pifdecavirus, and a further group of four—three clonal and one diverged—may represent a novel species within an as-yet-undefined genus most closely related to Pseudomonas phage QdP_Psa3. The most striking discovery was phage φCB10, a jumbo phage with a genome of approximately 310 kilobases encoding 476 predicted coding sequences, most of them of unknown function. φCB10 was unrelated to the other lineages, most closely resembled Pseudomonas jumbo phage Psa21 at only 53.2 percent genomic similarity, and may represent the founding member of a novel genus. All twelve candidates were predicted to be obligately lytic, and bioinformatic screening found no lysogeny-associated genes, antimicrobial resistance determinants or metal resistance cassettes—characteristics considered indispensable for safe agricultural biocontrol.

Environmental stability assays revealed considerable heterogeneity among isolates. Exposure to ultraviolet radiation for one hour reduced viable titres of all phages, with the severity depending on wavelength: under UV-A at 365 nanometres, relevant to field conditions, only φCB16 fell below the detection limit, whereas UV-C at 254 nanometres drove the titres of most phages below detection. Freezing at −20 degrees Celsius produced a particularly polarizing split along predicted morphology: the three myovirus-like phages lost roughly 1.6 to 3.4 orders of magnitude in titre within a single day and became undetectable after 30 days, while podovirus-like isolates fared substantially better, with φCB02 and φCB05 showing no significant losses over the full month. This partitioning suggests that virion structural properties may broadly influence environmental persistence, a phenomenon the authors say warrants mechanistic investigation. Notably, some isolates appeared more stable at 22 degrees Celsius than at the conventional 4 degrees Celsius storage temperature, raising questions about the universality of phage storage conventions. Prolonged exposure to highly acidic conditions, potentially relevant given that blueberries thrive in soil at pH 4.0 to 5.5, was among the most consistently destabilizing stresses, although most phages retained functionality over biologically relevant timescales.

To quantify antibacterial activity, the team performed planktonic killing assays against eight virulent bacterial strains, summarizing 64 phage-host pairs with a Growth Reduction Coefficient derived from bacterial growth curves. Individual phage performance varied widely and depended strongly on the strain: the jumbo phage φCB10 achieved the highest median coefficient, 0.943, and was the only phage rated as strong across all targeted hosts, while roughly half of all single-phage interactions were deemed poor. The two most effective phages were variously combined with the four least effective to construct eight cocktails of three to five variants. Although cocktails did not significantly outperform individual phages overall, four formulations—φC_A, φC_B, φC_C and φC_D—achieved median coefficients above 0.75 and were rated as strong across the full host panel, and were carried forward to plant-based testing.

The decisive validation used a detached-leaf model in which young Draper-variety leaves were pretreated with phage cocktails and then inoculated with two highly virulent strains. Bacterial densities were substantially reduced in all pretreated leaves. At early timepoints corresponding to maximal bacterial densities in untreated leaves, the four cocktails induced reductions of 4.5 to 5.5 logs against strain B11 and 3.2 to 5.6 logs against strain B04. At the ten-day endpoint, two cocktails pushed B11 below the detection limit, and two others reduced B04 below detection. A variant of the Growth Reduction Coefficient calculated from bacterial density over time rated all four cocktails as strong against B11, with φC_D also strong against B04. Critically, ex planta performance correlated strongly with in vitro performance, suggesting that planktonic killing assays can serve as a useful proxy for more complex validation models in high-throughput screens.

Measurements of phage densities inside the leaves supported the conclusion that bacterial suppression was driven by ongoing lytic replication rather than a single contact-mediated kill. The mean in situ phage-to-bacterium ratio rose rapidly to a peak of roughly 1.8 × 10⁴ for treatments targeting B11 before decaying by day ten, and endpoint phage-to-bacterium ratios correlated positively with overall treatment efficacy. The authors interpret this as evidence that successful biocontrol depends not only on the killing efficiency of individual phages but also on their capacity to persist within the plant environment and maintain productive infections over extended periods.

The study establishes a framework for the systematic identification and evaluation of candidate phages for agricultural biocontrol, while flagging the questions that remain before field deployment. Detached-leaf assays cannot fully recapitulate whole-plant systems, and the authors emphasize the need to reproduce these results in larger-scale studies with intact plants and, ultimately, under field conditions, alongside assessments of any impacts on plant health or the native phyllosphere microbiome. The mechanisms governing phage susceptibility among bacterial strains also remain poorly understood. Meanwhile, φCB10—with its exceptional antibacterial performance, favourable stability, enormous coding capacity and potential to carry accessory functions capable of circumventing bacterial defence systems such as restriction-modification and CRISPR-Cas—stands out as a particularly compelling subject for future investigation, both as a biocontrol candidate and as a window into the biology of underexplored jumbo phages.

Subject of Research: Isolation and evaluation of bacteriophages for biocontrol of Pseudomonas syringae complex populations on highbush blueberry leaves

Article Title: Bacteriophages Control Epiphytic Pseudomonas syringae Populations in Highbush Blueberry Leaves

Article References: Ball, C., Lauman, P., Xu, T., Guy, T., Dadej, K., Richter, R., Lubberts, M., Cross, K., Ren, M., Latchman, S. R., Burlakoti, R., Deng, X., Fong, K., & Wang, S. (2026). Bacteriophages Control Epiphytic Pseudomonas syringae Populations in Highbush Blueberry Leaves. Microbial Biotechnology, 19(9), Article e70438. https://doi.org/10.1111/1751-7915.70438

Image Credits: AI Generated

DOI: 10.1111/1751-7915.70438

Keywords: bacteriophages, phage biocontrol, Pseudomonas syringae complex, highbush blueberry, Vaccinium corymbosum, plant pathology, epiphytic bacteria, jumbo phage, phage cocktails, agricultural biotechnology, British Columbia, leaf blight