Hi everyone,

It’s time for another Phage Picks, a new monthly format for highlighting our favorite papers.

If we missed a cool or important paper — email us to let us know: [email protected].

Nanopore and Illumina sequencing reveal different viral populations from human gut samples

What is it about?

Evelien’s team compared Illumina’s short-read sequencing against ONT’s long-read sequencing, and various assembly strategies for getting phage genomes from fecal samples. It turns out that Illumina is generally better for recovering fully resolved viral genomes, while ONT is better for capturing broader viral diversity. The paper then discusses using hybrid approaches to get more authentic representations of viral diversity.

From Evelien’s tweet: “Sequencing platform, assembler, binning tools all recover different viral populations and have their pros and cons, choose wisely based on your research question.”

Why I’m excited about it:

Aside from the findings, the papers goes quite deeply into mixing, matching, and using various assembly and binning tools.

This is the kind of paper that might be very useful for “saving for later” if you ever find yourself in the depths of figuring out “when and how should I use Illumina vs. ONT… with which tools, and using what settings and configurations?”

~ Jan

Access: https://doi.org/10.1099/mgen.0.001236

Cook, R., Telatin, A., Hsieh, S., Newberry, F., Tariq, M. A., Baker, D. J., Carding, S. R., & Adriaenssens, E. M. (2024). Nanopore and Illumina sequencing reveal different viral populations from human gut samples. Microbial Genomics, 10(4). https://doi.org/10.1099/mgen.0.001236

Prediction of Klebsiella phage-host specificity at the strain level

What is it about?

The authors built a predictive machine learning model for Klebsiella phage-host infectivity. They used a combination of ESM-2 protein language model for creating “vector” representations of receptor-binding proteins and Klebsiella-specific capsule polysaccharides called K-antigens. They then created an XGBoost model for predicting if a provided pair of phages and bacteria can lead to infectivity.

Why I’m excited about it:

I love how this paper spells out how machine learning can be used predict to phage-host infectivity. The authors go into lengths to both share the data and explain / document the code. The code is fully documented, and even includes an easy to use, step by step Jupyter Notebook to help readers replicate their results.

Though the model is very specific to RBPs and K-antigens, the paper and accompanying code acts like a tutorial for other labs to follow suit and build their own models predictive.

I love clearly the laid out all the steps and all the code — and anyone with a M1 Macbook (or faster) can reproduce all the results, and even make their own (Klebsiella strain) phage-host predictions with the accompanying notebook.

Overall, it’ll hopefully help others create more diagnostic tools across all of the ESKAPE pathogens that severely cut down the time it takes to find suitable phages.

~ Jan

Twitter thread: https://twitter.com/dimiboeckaerts/status/1793321221846880599

Paper: https://doi.org/10.1038/s41467-024-48675-6

Github: https://github.com/dimiboeckaerts/PhageHostLearn

Data: https://zenodo.org/records/11061100

Boeckaerts, D., Stock, M., Ferriol-González, C., Oteo-Iglesias, J., Sanjuán, R., Domingo-Calap, P., De Baets, B., & Briers, Y. (2024). Prediction of Klebsiella phage-host specificity at the strain level. In Nature Communications (Vol. 15, Issue 1). Springer Science and Business Media LLC. https://doi.org/10.1038/s41467-024-48675-6

Resistance to bacteriophage incurs a cost to virulence in drug-resistant Acinetobacter baumannii

What is it about?

The authors isolated two phages — LemonAid and Tonic — that can infect drug-resistant A. baumannii. They also discovered that the A. baumannii developed resistance to these phages at the cost to virulence, correlated with a gene potentially affection a phage receptor.

Why I’m excited about it:

These phage resistance / virulence / antibiotic resistance tradeoff papers are super interesting to me. These kinds of papers show that we can potentially drive phages to “evolve” in certain directions by giving up certain characteristics, like virulence. Similar to Yale’s efflux pump-affecting phage, OMKO-1, it shows that we need to be collecting phages based on how we know the phage will affect the bacterial target (whether it’s through lysis or merely driving it to evolve in a certain way, leading it to an evolutionary trap where we can then kill it with antibiotics). The paper also highlights the necessity to discover and document more phage-host interactions that change either the phage or host characteristics.

~ Jan

Paper: https://doi.org/10.1099/jmm.0.001829

Manley, R., Fitch, C., Francis, V., Temperton, I., Turner, D., Fletcher, J., Phil, M., Michell, S., & Temperton, B. (2024). Resistance to bacteriophage incurs a cost to virulence in drug-resistant Acinetobacter baumannii. In Journal of Medical Microbiology (Vol. 73, Issue 5). Microbiology Society. https://doi.org/10.1099/jmm.0.001829

Temperate phage-antibiotic synergy across antibiotic classes reveals new mechanism for preventing lysogeny

What is it about?

The author of this paper discovered that various antibiotics can induce lysis from otherwise temperate phages. Surprisingly, some of the antibiotics that worked were not traditionally known to induce an “SOS” response. A novel protein synthesis inhibitor mechanism seems to be the culprit behind this antibiotic-phage synergy behavior.

Why I’m excited about it:

It seems like there are so many more mechanisms and relationships out there we don’t know about, between phages, bacteria, and antibiotics. This paper highlights the need to check how antibiotics will affect lysis (including temperate phages), the necessity to identify the mechanisms, and the need to systematically collect data, even in situations where “we obviously know it won’t work” — because we actually know so little about these phage / host / antibiotic interactions.

It also highlights the need for more larger scale data collection efforts to help us chart out new mechanisms that affect lysis.

Lastly, I’m excited to combine ideas from this (and similar) papers to the idea of deep model predictive control — the notion that we can use a machine learning model that can predict the outcomes of phage/host/antibiotic combinations to “direct” the evolution based on recommendations from the model, like in this paper, where they directed gene expression via an optogenetic system.

~ Jan

Access: https://doi.org/10.1128/mbio.00504-24

Al-Anany, A. M., Fatima, R., Nair, G., Mayol, J. T., & Hynes, A. P. (2024). Temperate phage-antibiotic synergy across antibiotic classes reveals new mechanism for preventing lysogeny. In V. Sperandio & A. Maresso (Eds.), mBio. American Society for Microbiology. https://doi.org/10.1128/mbio.00504-24

I hope you enjoyed this month’s Picks! There’s too many cool papers coming out, especially at the intersection of phage, bioinformatics and machine learning (as you can see, I’m quite biased…). I’ll let Jess pick some papers next time…

See you next month!
~ Jan