Rafal Mostowy
@rafalmostowy.bsky.social
Group leader at Jagiellonian University in Krakow, cofounder of Polonium Foundation. Working on evolution of sex in bugs.
This underscores the complexity of predicting enzyme specificity — even similar proteins may target different capsules.
Yet, the study reveals an enormous functional and structural diversity encoded in temperate phages.
Yet, the study reveals an enormous functional and structural diversity encoded in temperate phages.
June 26, 2025 at 7:13 AM
This underscores the complexity of predicting enzyme specificity — even similar proteins may target different capsules.
Yet, the study reveals an enormous functional and structural diversity encoded in temperate phages.
Yet, the study reveals an enormous functional and structural diversity encoded in temperate phages.
But predictions are just the start.
Together with the group of Zuzanna Drulis-Kawa, we experimentally tested 50 candidate enzymes from prophages on a panel of 119 capsule types.
Only 14 were active — and not always on the predicted target.
🧪🧫
Together with the group of Zuzanna Drulis-Kawa, we experimentally tested 50 candidate enzymes from prophages on a panel of 119 capsule types.
Only 14 were active — and not always on the predicted target.
🧪🧫
June 26, 2025 at 7:13 AM
But predictions are just the start.
Together with the group of Zuzanna Drulis-Kawa, we experimentally tested 50 candidate enzymes from prophages on a panel of 119 capsule types.
Only 14 were active — and not always on the predicted target.
🧪🧫
Together with the group of Zuzanna Drulis-Kawa, we experimentally tested 50 candidate enzymes from prophages on a panel of 119 capsule types.
Only 14 were active — and not always on the predicted target.
🧪🧫
We discovered a widespread class of SGNH hydrolase–containing RBPs — enzymes structurally related to deacetylases.
They likely modify capsules (e.g. by removing acetyl groups) instead of degrading them — a very different phage infection strategy!
They likely modify capsules (e.g. by removing acetyl groups) instead of degrading them — a very different phage infection strategy!
June 26, 2025 at 7:13 AM
We discovered a widespread class of SGNH hydrolase–containing RBPs — enzymes structurally related to deacetylases.
They likely modify capsules (e.g. by removing acetyl groups) instead of degrading them — a very different phage infection strategy!
They likely modify capsules (e.g. by removing acetyl groups) instead of degrading them — a very different phage infection strategy!
We found 16 capsule types (K-loci) with strong genetic predictors, mostly receptor-binding proteins (RBPs).
Some were classical depolymerases… but many were not.
Some were classical depolymerases… but many were not.
June 26, 2025 at 7:13 AM
We found 16 capsule types (K-loci) with strong genetic predictors, mostly receptor-binding proteins (RBPs).
Some were classical depolymerases… but many were not.
Some were classical depolymerases… but many were not.
But...
temperate phages (those that integrate into the bacterial genome) are less understood.
We analysed 3,900 Klebsiella genomes and identified over 8,100 prophages, then applied GWAS to link phage genes to capsule types.
temperate phages (those that integrate into the bacterial genome) are less understood.
We analysed 3,900 Klebsiella genomes and identified over 8,100 prophages, then applied GWAS to link phage genes to capsule types.
June 26, 2025 at 7:13 AM
But...
temperate phages (those that integrate into the bacterial genome) are less understood.
We analysed 3,900 Klebsiella genomes and identified over 8,100 prophages, then applied GWAS to link phage genes to capsule types.
temperate phages (those that integrate into the bacterial genome) are less understood.
We analysed 3,900 Klebsiella genomes and identified over 8,100 prophages, then applied GWAS to link phage genes to capsule types.
Phages use surface-binding proteins to recognise their bacterial hosts.
In Klebsiella, this often means targeting the capsule — a sugary layer that varies across strains. Some phages bring depolymerases, enzymes that slice through these polysaccharides to enable infection.
In Klebsiella, this often means targeting the capsule — a sugary layer that varies across strains. Some phages bring depolymerases, enzymes that slice through these polysaccharides to enable infection.
June 26, 2025 at 7:13 AM
Phages use surface-binding proteins to recognise their bacterial hosts.
In Klebsiella, this often means targeting the capsule — a sugary layer that varies across strains. Some phages bring depolymerases, enzymes that slice through these polysaccharides to enable infection.
In Klebsiella, this often means targeting the capsule — a sugary layer that varies across strains. Some phages bring depolymerases, enzymes that slice through these polysaccharides to enable infection.
What determines who a phage can infect?
We tackled this question for temperate phages of Klebsiella — a bacterial pathogen — using a genome-wide association study (GWAS) and a massive protein testing effort.
👇 A thread!
We tackled this question for temperate phages of Klebsiella — a bacterial pathogen — using a genome-wide association study (GWAS) and a massive protein testing effort.
👇 A thread!
June 26, 2025 at 7:13 AM
What determines who a phage can infect?
We tackled this question for temperate phages of Klebsiella — a bacterial pathogen — using a genome-wide association study (GWAS) and a massive protein testing effort.
👇 A thread!
We tackled this question for temperate phages of Klebsiella — a bacterial pathogen — using a genome-wide association study (GWAS) and a massive protein testing effort.
👇 A thread!
I'm looking for a postdoc in computational evolutionary phage genomics, with the goal to better understand how protein innovation arises in viruses, at @jagiellonskiuni.bsky.social in Krakow (Poland).
Drop me a line if it sounds cool!
Deadline: 20.06.
Link here:
mcb.uj.edu.pl/documents/15...
Drop me a line if it sounds cool!
Deadline: 20.06.
Link here:
mcb.uj.edu.pl/documents/15...
May 20, 2025 at 7:52 AM
I'm looking for a postdoc in computational evolutionary phage genomics, with the goal to better understand how protein innovation arises in viruses, at @jagiellonskiuni.bsky.social in Krakow (Poland).
Drop me a line if it sounds cool!
Deadline: 20.06.
Link here:
mcb.uj.edu.pl/documents/15...
Drop me a line if it sounds cool!
Deadline: 20.06.
Link here:
mcb.uj.edu.pl/documents/15...
With MANIAC we also obtained some insights into ANI-based taxonomic assignments of dsDNA phages: we found that while it works well for lytic phages, it struggles with temperate phages—likely due to their extensive recombination & mosaicism, as seen in λ-like phages.
January 30, 2025 at 9:50 AM
With MANIAC we also obtained some insights into ANI-based taxonomic assignments of dsDNA phages: we found that while it works well for lytic phages, it struggles with temperate phages—likely due to their extensive recombination & mosaicism, as seen in λ-like phages.
To me, this bears resemblance to a recombination threshold hypothesis in bacteria, where above this threshold recombination maintains species cohesiveness (high AF), and below it enhances diversification and evolution (low AF). But it's a hypothesis at this stage. (Figure from PMID 19197054)
January 30, 2025 at 9:50 AM
To me, this bears resemblance to a recombination threshold hypothesis in bacteria, where above this threshold recombination maintains species cohesiveness (high AF), and below it enhances diversification and evolution (low AF). But it's a hypothesis at this stage. (Figure from PMID 19197054)
Using MANIAC, we noticed that ANI distribution in dsDNA phages shows a distinct gap around 80%. Above this, genomes align well; below it, AF drops. This pattern holds even in large, metagenome-derived datasets like PhageScope, though there we see many more recent recombinants (high ANI and low AF).
January 30, 2025 at 9:50 AM
Using MANIAC, we noticed that ANI distribution in dsDNA phages shows a distinct gap around 80%. Above this, genomes align well; below it, AF drops. This pattern holds even in large, metagenome-derived datasets like PhageScope, though there we see many more recent recombinants (high ANI and low AF).
MANIAC isn’t just accurate—it’s fast. Its "Fast" setting, which is still reliably accurate for ANI~70%, can process RefSeq on a laptop within minutes and can handle metagenomic datasets on HPCs.
January 30, 2025 at 9:50 AM
MANIAC isn’t just accurate—it’s fast. Its "Fast" setting, which is still reliably accurate for ANI~70%, can process RefSeq on a laptop within minutes and can handle metagenomic datasets on HPCs.
We benchmarked MANIAC's accuracy against both BLAST-based & simulated datasets, showing it estimates ANI reliably over a wider range of sequence divergence than existing tools like CheckV, FastANI or MUMmer, which are better suited for the ANI>80% range.
January 30, 2025 at 9:50 AM
We benchmarked MANIAC's accuracy against both BLAST-based & simulated datasets, showing it estimates ANI reliably over a wider range of sequence divergence than existing tools like CheckV, FastANI or MUMmer, which are better suited for the ANI>80% range.
Most ANI tools handle closely related genomes well (>80% ANI). But for viruses, we often care about ANI around 70%—a range where alignment can be tricky due to genome diversity & rearrangements. MANIAC tackles this using a fragment-based approach inspired by bacterial genome comparisons.
January 30, 2025 at 9:50 AM
Most ANI tools handle closely related genomes well (>80% ANI). But for viruses, we often care about ANI around 70%—a range where alignment can be tricky due to genome diversity & rearrangements. MANIAC tackles this using a fragment-based approach inspired by bacterial genome comparisons.
🚀 Follow-up on my recent MANIAC tweet: our new MMseqs2-blased tool is optimised for ANI and alignment fraction (AF) calculation in viral genomes. Unlike many other tools, it excels at ANI ~70%, an important range for viral taxonomy & evolution.
Here’s why that matters ➡️ a thread! 🧵
Here’s why that matters ➡️ a thread! 🧵
January 30, 2025 at 9:50 AM
🚀 Follow-up on my recent MANIAC tweet: our new MMseqs2-blased tool is optimised for ANI and alignment fraction (AF) calculation in viral genomes. Unlike many other tools, it excels at ANI ~70%, an important range for viral taxonomy & evolution.
Here’s why that matters ➡️ a thread! 🧵
Here’s why that matters ➡️ a thread! 🧵
Our new paper out in mSystems!
journals.asm.org/eprint/ISRAZ...
We created a tool called MANIAC to estimate ANI between virus pairs. We found that:
1) MANIAC can reliably estimate ANI ~70%,
2) dsDNA phages exhibit ANI gap around 80%,
3) taxonomic assignment in temperate phages is tricky.
Try it!
journals.asm.org/eprint/ISRAZ...
We created a tool called MANIAC to estimate ANI between virus pairs. We found that:
1) MANIAC can reliably estimate ANI ~70%,
2) dsDNA phages exhibit ANI gap around 80%,
3) taxonomic assignment in temperate phages is tricky.
Try it!
January 29, 2025 at 2:26 PM
Our new paper out in mSystems!
journals.asm.org/eprint/ISRAZ...
We created a tool called MANIAC to estimate ANI between virus pairs. We found that:
1) MANIAC can reliably estimate ANI ~70%,
2) dsDNA phages exhibit ANI gap around 80%,
3) taxonomic assignment in temperate phages is tricky.
Try it!
journals.asm.org/eprint/ISRAZ...
We created a tool called MANIAC to estimate ANI between virus pairs. We found that:
1) MANIAC can reliably estimate ANI ~70%,
2) dsDNA phages exhibit ANI gap around 80%,
3) taxonomic assignment in temperate phages is tricky.
Try it!
We then used sequence information to find more cases of recently emerged domain mosaicism. We saw examples of domain swap and shuffling not only in receptor-binding proteins and endolysins, but also in replication proteins like ribonucleotide reductases or replication initiation proteins. 5/7
November 28, 2023 at 1:55 PM
We then used sequence information to find more cases of recently emerged domain mosaicism. We saw examples of domain swap and shuffling not only in receptor-binding proteins and endolysins, but also in replication proteins like ribonucleotide reductases or replication initiation proteins. 5/7
Some PHROG classes were highly mosaic with cases of contemporary mosaicism (similar protein sequence of the shared fragment): receptor-binding proteins (tail fibres and tail spikes), endolysins and DNA polymerases. 4/7
November 28, 2023 at 1:54 PM
Some PHROG classes were highly mosaic with cases of contemporary mosaicism (similar protein sequence of the shared fragment): receptor-binding proteins (tail fibres and tail spikes), endolysins and DNA polymerases. 4/7
We found that such mosaicism is extensive, with proteins of diverse functions sharing domains. Many such cases were due to ancient mosaicism, eg via sharing conserved domains like Ig-like, P-loop related or ribonuclease H-like. However, for some functions this mosaicism wasn't ancient. 3/7
November 28, 2023 at 1:54 PM
We found that such mosaicism is extensive, with proteins of diverse functions sharing domains. Many such cases were due to ancient mosaicism, eg via sharing conserved domains like Ig-like, P-loop related or ribonuclease H-like. However, for some functions this mosaicism wasn't ancient. 3/7
We used a sensitive HMM-HMM approach to do all-by-all comparison of 134k representative phage proteins from NCBI RefSeq phages. We then predicted their functions using PHROGs and domains using ECOD. We looked for examples of domain mosaicism, i.e. domain sharing between non-homologous proteins. 2/7
November 28, 2023 at 1:52 PM
We used a sensitive HMM-HMM approach to do all-by-all comparison of 134k representative phage proteins from NCBI RefSeq phages. We then predicted their functions using PHROGs and domains using ECOD. We looked for examples of domain mosaicism, i.e. domain sharing between non-homologous proteins. 2/7
Our group had a great time at the Local Pangenome conference in Alicante. Very proud of Janusz and Bogna for giving great talks.
October 28, 2023 at 8:42 PM
Our group had a great time at the Local Pangenome conference in Alicante. Very proud of Janusz and Bogna for giving great talks.