🦠 microbiology

PhagePickr: A bacteria-centric computational tool for designing resistance-aware phage cocktails

PhagePickr is a novel, bacteria-centric computational pipeline that designs resistance-aware phage cocktails by selecting phages based on bacterial surface receptor similarity and genomic diversity to minimize the risk of host resistance evolution.

Original authors: Oneto, A., Okamoto, K. W.

Published 2026-10-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Oneto, A., Okamoto, K. W.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

For decades, the medical world has relied on antibiotics to fight bacterial infections, but bacteria have learned to outsmart these drugs, evolving into superbugs that no longer respond to treatment. This growing resistance has created a desperate need for new ways to kill harmful bacteria. One promising alternative is phage therapy, which uses viruses called bacteriophages that naturally hunt and destroy bacteria. However, just as bacteria can evolve resistance to antibiotics, they can also evolve defenses against these viruses. When a bacterium changes its surface to block a virus, that specific virus can no longer infect it, rendering the treatment useless. The challenge for scientists is to design a treatment that stays one step ahead, preventing the bacteria from building a wall that stops the attack.

To solve this, researchers Alessandro Oneto and Kenichi W. Okamoto have developed a new computer tool called PhagePickr. Instead of trying to guess which viruses might work, this tool uses a strategy based on the specific "doors" bacteria use to let viruses in. Bacteria have proteins on their surfaces that act as receptors, or entry points, for viruses. If a virus can recognize and bind to these receptors, it can infect the cell. The researchers realized that if they could find bacteria that have very similar surface doors to the dangerous bacteria they want to treat, they might find viruses that are already adapted to those doors. By selecting a mix of viruses that are genetically very different from one another but all capable of targeting similar entry points, the tool aims to create a cocktail that is much harder for bacteria to resist. If a bacterium tries to change one door to escape one virus, it might still be vulnerable to the others in the mix, which use different strategies to get inside.

The team tested their new software on a group of notorious, drug-resistant bacteria known as the ESKAPE pathogens, which include species like Staphylococcus aureus and Pseudomonas aeruginosa, as well as two less-studied species. The computer program worked by first looking at the surface proteins of the target bacteria and then searching a database to find other bacterial species that shared a very similar set of these proteins. Once it identified these "neighbors," the tool looked at the viruses known to infect those neighbors. It then selected the most genetically diverse pairs of viruses from each group to ensure the final mix was as varied as possible. This approach allowed the tool to suggest treatments that included not only viruses known to infect the target directly, but also viruses from related species that might be effective due to their shared surface features.

The results showed that PhagePickr could successfully generate these viral cocktails for all the tested bacteria. For the well-known drug-resistant bacteria, the tool often found viruses that had already been shown in other studies to be effective, confirming that the method works. For the less-studied species, the tool proposed new candidates, including viruses from distant bacterial families that shared the necessary surface traits. In some cases, the tool suggested viruses that were known to be lytic, meaning they burst and kill the bacteria immediately, while in other cases, it identified viruses that might need further study to confirm their behavior. The researchers noted that while the computer could build these lists of potential candidates, the actual ability of these viruses to cure an infection in a living patient has not yet been tested. The tool is designed to be a starting point, offering a shortlist of the most promising viruses for scientists to test in a lab.

One of the key findings was that the tool could handle bacteria with very few known viruses by looking at their closest relatives. For example, when analyzing a difficult-to-study bacterium, the software found a neighbor species that was genetically distant but shared the same surface receptors, allowing it to suggest viruses that might work. The researchers also found that the time it took to run the program depended mostly on the size of the viral genomes being compared, with larger sets of data taking longer to process. While the tool is powerful, the authors are careful to point out that it is currently a hypothesis generator. It relies on the quality of the data available in public databases, and if the information about bacterial surface proteins is incomplete, the suggestions might be less precise. As more data becomes available about how bacteria and viruses interact, the tool is expected to become even more accurate.

Ultimately, PhagePickr represents a shift in how scientists approach the design of viral therapies. Instead of relying on trial and error or looking only at the most obvious matches, this tool uses a systematic, data-driven approach to find hidden connections between bacteria and viruses. By focusing on the shared physical traits of the bacteria rather than just their family tree, it opens up a wider range of potential treatments. The researchers hope that by providing a clear, automated way to select diverse viral candidates, they can help speed up the discovery of new therapies that can outmaneuver the evolving defenses of drug-resistant bacteria. The work suggests that with better data and continued testing, this kind of computational guidance could become a standard part of the fight against superbugs.

Technical Summary: PhagePickr

Problem Statement
Antibiotic resistance poses a critical global health threat, necessitating alternatives such as bacteriophage (phage) therapy. However, bacteria rapidly evolve resistance to phages, often by modifying or masking surface receptors, which limits the efficacy of single-phage treatments. While phage cocktails (combinations of multiple phages) are a proposed solution to delay resistance, current computational tools for cocktail design (e.g., PhageCocktail) focus on infection networks or genomic similarity but fail to explicitly address the evolution of bacterial resistance. Specifically, existing tools do not leverage bacterial surface receptor profiling to select phages that target diverse receptors, a strategy hypothesized to make simultaneous resistance evolution more difficult for bacteria.

Methodology
The authors present PhagePickr, a computational pipeline designed to automate the design of resistance-aware phage cocktails using a "bacteria-centric" approach. The workflow operates in two primary stages:

Receptor Profiling and Nearest-Neighbor Identification:
- PhagePickr utilizes a built-in dataset of 200 pathogenic species, constructed from NCBI's Identical Protein Groups (IPG) and Nucleotide databases.
- It constructs a binary presence/absence matrix ( $A$ ) of 2,951 receptor-associated protein features across these species. Features are derived from NCBI record titles containing the term "receptor" in any field.
- Given a target bacterial species, the tool identifies phenotypically similar hosts by calculating the normalized Hamming distance between the target's receptor feature vector and those of other species in the database.
- The $n$ closest neighbors (including the target species by default) are selected. The underlying hypothesis is that species with similar receptor configurations may share susceptibility to phages, even if they are phylogenetically distant.
Phage Selection via Genomic Divergence:
- For each of the $n$ selected hosts, the tool retrieves registered phage genomes.
- To maximize the likelihood of targeting different receptors and reducing cross-resistance, PhagePickr prioritizes genomic diversity.
- For hosts with more than two registered phages, genomes are aligned using MAFFT. A pairwise identity distance matrix is computed, and the $k$ most divergent phage pairs are selected.
- Hosts with two or fewer phages contribute all available phages to the cocktail.
- The final cocktail is the aggregation of these selected phages, with a maximum theoretical size of $2nk$ (default parameters: $n=3$ , $k=1$ , yielding 3–6 phages).

Key Contributions

Novel Framework: PhagePickr is the first automated tool to design phage cocktails based explicitly on bacterial receptor similarity and phage genomic divergence to mitigate resistance evolution.
Data Integration: It integrates NCBI data into a reusable receptor feature matrix and phage-host association records, allowing for scalable updates as databases grow.
Experimental Features: The tool includes an optional, experimental module using BACPHLIP to predict phage lifestyles (lytic vs. temperate), allowing users to filter for lytic candidates, though this is not integrated into the default workflow.
Scalability: The pipeline is designed to run on standard UNIX/Windows (via WSL) systems with manageable memory usage, though runtime is bottlenecked by multiple sequence alignment (MSA) of phage genomes.

Results
The authors evaluated PhagePickr on eight targets: the six ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter spp.) and two understudied species (Burkholderia pseudomallei and Pasteurella multocida).

Cocktail Generation: The tool successfully generated cocktails of 4 to 6 phages for all targets.
Neighbor Selection: The nearest-neighbor algorithm identified hosts ranging from closely related species (e.g., S. aureus neighbors within the Staphylococcus genus) to phylogenetically distant genera (e.g., Clostridium and Staphylococcus for E. faecium), driven by shared receptor feature profiles.
Diversity Maximization: Comparative analysis showed that cocktails selected by PhagePickr possessed significantly higher genomic diversity scores (average maximum pairwise identity distance) than random controls, confirming the efficacy of the selection step in maximizing divergence.
Therapeutic Potential: Manual evaluation revealed that most cocktails contained at least one phage with reported lytic activity against the target or its neighbors. However, many candidates lacked comprehensive experimental validation regarding infectivity, lifestyle, or host range.
Performance: Runtime was heavily dependent on the total length of aligned phage sequences (scaling with a power law of ~1.54), with the most time-consuming task being the MSA step. Memory usage remained low (<1.5 GB) across all tests.

Significance and Claims
The paper positions PhagePickr as a hypothesis-generating tool rather than a clinically validated solution. Its primary significance lies in providing a systematic, reproducible framework for "resistance-aware" cocktail design that leverages receptor phenotypes to predict cross-infectivity.

The authors explicitly state that:

The generated cocktails have not been experimentally validated for therapeutic efficacy against the specific target strains.
The tool's utility is currently limited by the depth and accuracy of receptor annotations in public databases; sparse annotations (e.g., for E. faecium) lead to exploratory, cross-genus selections that require further characterization.
The approach represents a key step toward rational design, aiming to reduce the risk of cross-resistance by selecting genomically divergent phages that may utilize different infection pathways.
Future iterations could incorporate strain-level comparisons, antibiotic-resistance receptor targeting to induce collateral sensitivity, and more robust lifestyle filtering, contingent on improved genomic and phenotypic data availability.

In summary, PhagePickr offers a computational strategy to bridge the gap between bacterial receptor biology and phage therapy design, aiming to preempt resistance evolution through diversity, while acknowledging that experimental validation remains the necessary next step for any proposed cocktail.

Technical Summary: PhagePickr

More like this