Sign in

Daniel Krentzel

@danielkrentzel.bsky.social
206 followers 305 following 56 posts

Postdoc-ing in @virginieuhlmann.bsky.social's lab at UZH 🇨🇭 | Previously @pasteur.fr, @crick.ac.uk and @imperialcollegeldn.bsky.social

PostsRepliesMedia
Reposted by Daniel Krentzel
TheLazarLab @thelazarlab.bsky.social · 02/10/2026
Counting bacteria is not enough. To understand what makes a bacterium dangerous, we need to see what it actually does when it meets human cells. Our MALVINA study is now out in @natcomms.nature.com www.nature.com/articles/s41...
nature.com
Deep-learning single-cell profiling uncovers interbacterial and drug-driven reshaping of pathogen virulence - Nature Communications
This study introduces MALVINA, a deep-learning platform that measures bacterial invasion and DNA damage in individual human cells, revealing how bacterial interactions and commonly used drugs reshape ...
153
Reposted by Daniel Krentzel
Virginie Uhlmann @virginieuhlmann.bsky.social · 05/10/2026
Are you a biologist working with microscopy images? A computer scientist wanting to work with microscopy data? Their boss??? None of these, but curious how AI has transformed microscopy image analysis? 🔬🤖 This one's for you: 📄 doi.org/10.1146/annu... More in the thread! 🧵
doi.org
From Pixels to Discovery: Microscopy Image Analysis in the Age of Artificial Intelligence
Modern microscopy generates data volumes that far exceed human analytical capacity. Over the past decade, artificial intelligence (AI) has transformed microscopy image analysis into an engine for quan...
17128
Reposted by Daniel Krentzel
Jean-Yves Tinevez @jytinevez.bsky.social · 05/10/2026
We have a position open in our core facility in the @pasteur.fr . Please share it and / or apply! forum.image.sc/t/research-e...
forum.image.sc
Research engineer in Bioimage Analysis. Institut Pasteur, Image Analysis Hub - 2026
Dear all Below is an announcement for a permanent position as a Research Engineer in the Institut Pasteur, Paris. There is a flyer attached at the bottom. We are looking ideally for a Bioimage An...
15276
Reposted by Daniel Krentzel
Johannes Soltwedel @jo-soltwedel.bsky.social · 01/10/2026
Just released the napari-clusters-plotter 0.11.0 github.com/BiAPoL/napar..., a little passion-project of mine for interactive exploration of datasets and attached features in @napari.org Two cool things to take note of 👇
github.com
GitHub - BiAPoL/napari-clusters-plotter: A napari plugin for clustering objects according to their properties.
A napari plugin for clustering objects according to their properties. - BiAPoL/napari-clusters-plotter
13113
Daniel Krentzel @danielkrentzel.bsky.social · 01/10/2026
Thank you!! 🙏
010
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
Massive thank you to the wonderful team behind this project! 🙏 @awehenkel.bsky.social @spetrella.bsky.social @chzimmer.bsky.social @curieuseny.bsky.social
021
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
The source code and datasets for this study can be found here 👇 💻 github.com/krentzd/morp... 🔬 www.ebi.ac.uk/biostudies/b...
github.com
GitHub - krentzd/morphoscreen: Source code for morphoscreen analysis
Source code for morphoscreen analysis. Contribute to krentzd/morphoscreen development by creating an account on GitHub.
112
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
5️⃣ Finally, we asked whether we could use our model to provide insights into fundamental biological processes by studying the bacterial cell cycle of C. glutamicum. We show that we can recover known mechanistic relationships between bacterial genes directly from images of mutants.
100
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
4️⃣ We then show that we can use the feature vectors of our deep learning model to link images of drug-treated bacteria to mutants in which the same or a similar pathway has been disrupted. We believe that this result is an important step towards mutant-based target identification of hit compounds.
110
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
3️⃣ Next, we ensured that our model could identify the mode of action of drugs that were excluded from its training data. And when we applied hierarchical clustering to the feature vectors of our model, we found that images of drug-treated bacteria cluster according to which pathways the drugs target.
100
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
2️⃣ We then verified that our deep learning model was able to robustly recognise the mode of action of these reference drugs on hold-out test replicates.
100
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
1️⃣ We started by collecting a dataset of C. glutamicum bacteria exposed to a set of reference TB drugs and then trained a neural network with contrastive learning to identify the mode of action of these drugs directly from images.
100
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
Working with Mtb (the causative agent of TB) is challenging and can be dangerous. This is why we developed our method around a well-studied and non-pathogenic surrogate model called C. glutamicum which is normally used to produce umami flavouring 🍜 www.ajinomoto.com.vn/en/what-is-m...
100
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
TB is the leading cause of death from a single infectious agent with patients required to take up to 14,600 pills and endure 240 injections. And yet, the chance of survival is only ~50% for these patients. ℹ️ Infographic by @msfaccess.org 👇 🔗 msfaccess.org/infographic-...
100
Daniel Krentzel @danielkrentzel.bsky.social · 30/09/2026
Our paper outlining a deep learning-based phenotypic screening method to discover new TB drugs is out in @science.org #ScienceAdvances 💊🦠 It's the result of a great collaboration between the labs of @chzimmer.bsky.social and @awehenkel.bsky.social at @pasteur.fr! 📃 www.science.org/doi/10.1126/...
science.org
Deep learning extracts MoA-specific signatures from high-throughput images of chemically and genetically perturbed Corynebacteria
To address drug-resistant tuberculosis, the leading single-pathogen infectious killer, drugs with novel modes of action (MoAs) are urgently needed. Phenotypic screening of chemical libraries can ident...
2297
Reposted by Daniel Krentzel
Marie Held @makaefer.bsky.social · 28/09/2026
Want to enhance your image analysis skills with napari? We’re running a two-day training event on how to use napari to effectively analyse your bioimaging data. 🕒 Applications close on 15 September. Learn more and apply 👇 www.crick.ac.uk/whats-on/cbi...
crick.ac.uk
01216
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
And here's a nice résumé of our @natcomms.nature.com paper for the grand public 👇 📃 www.nature.com/articles/s41...
nature.com
Deep learning recognises antibiotic modes of action from brightfield images - Nature Communications
Authors present a deep learning model that recognises antibiotic modes of action (MoA) directly from brightfield images. Their proposed method complements traditional growth inhibition assays in the s...
092
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Thank you to all of the co-authors for coming along for the ride! 🙏 @kelvinkho.bsky.social, Julienne, Nassim, Thomas, Max, Agnès, @curieuseny.bsky.social, Spencer, Mark, @awehenkel.bsky.social, @ivo-boneca.bsky.social and @chzimmer.bsky.social
010
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
The source code and datasets used in this paper can be found here 👇 💻 github.com/krentzd/ai4ab 🔬 www.ebi.ac.uk/biostudies/b...
github.com
GitHub - krentzd/ai4ab: Source code for AI4AB project
Source code for AI4AB project. Contribute to krentzd/ai4ab development by creating an account on GitHub.
110
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
And finally, we show that our deep learning model also performs well on images of Klebsiella pneumoniae which is listed by the WHO as a bacterial priority pathogen.
110
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Next, we developed a mode-of-action novelty detection algorithm to detect if a previously unknown compound acts via a new mechanism. We believe that this could be crucial for rational hit selection of compounds.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Our method also identified the previously unknown mode of action of relebactam in E. coli which @kelvinkho.bsky.social confirmed through follow-up experiments in the wet lab.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Our method was able to robustly assign the mode of action of most compounds based on literature annotations. We did, however, observe that this was not the case for inhibitors of PBP1A/B. @kelvinkho.bsky.social then showed that all of the supposed PBP1A/B inhibitors in fact had different targets.
110
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
To test that our deep learning method would be able to correctly identify the mode of action of a compound that was not included in its training data, we designed a leave-one-compound-out experiment where we systematically removed one compound from the training data.
110
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
We used the distance of latent features to negative controls to detect if bacteria had been exposed to antibiotics and found that detection was successful even well below inhibitory concentrations.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Then, we looked at the latent representations of our deep learning model across different drug concentrations and saw that they moved away from negative controls as concentrations increased.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
We also found that classification performance at the highest concentration remained stable even when reducing the number of training images from 120 down to 8 per class!
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Next, we performed an ablation study to check if all of the imaging channels we used were actually required to obtain high classification accuracies. To our surprise, our deep learning model reached comparable compound and MoA classification performance even without any fluorescent labels.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Our deep learning model was able to both distinguish between compounds and different modes of action with very high accuracy, directly from images.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
We first acquired images of E. coli exposed to a set of reference antibiotics on a high-content screening system. Then, we trained a deep learning model to distinguish between treatment conditions directly from images.
100
Daniel Krentzel @danielkrentzel.bsky.social · 28/09/2026
Super excited to see the first research article from my PhD @pasteur.fr out in @natcomms.nature.com! Here's a thread on the key results 🧵 @chzimmer.bsky.social @kelvinkho.bsky.social @curieuseny.bsky.social @awehenkel.bsky.social @ivo-boneca.bsky.social www.nature.com/articles/s41...
nature.com
Deep learning recognises antibiotic modes of action from brightfield images - Nature Communications
Authors present a deep learning model that recognises antibiotic modes of action (MoA) directly from brightfield images. Their proposed method complements traditional growth inhibition assays in the s...
153
Reposted by Daniel Krentzel
Christophe Zimmer @chzimmer.bsky.social · 10/09/2026
Can you tell how an antibiotic kills bacteria just by watching them under a regular microscope — without any dyes ? Yes, it turns out, at least if you are a deep neural net.
12316
Reposted by Daniel Krentzel
Nature Communications @natcomms.nature.com · 17/09/2026
#Antibiotics A deep-learning model recognises antibiotic modes of action using unlabelled brightfield images of drug-treated bacteria @chzimmer.bsky.social @pasteur.fr @uni-wuerzburg.de @danielkrentzel.bsky.social @kelvinkho.bsky.social @awehenkel.bsky.social #deeplearning #AI4AMR #microbiology
dlvr.it
Deep learning recognises antibiotic modes of action from brightfield images - Nature Communications
Authors present a deep learning model that recognises antibiotic modes of action (MoA) directly from brightfield images. Their proposed method complements traditional growth inhibition assays in the search for novel antibiotics.
031
Reposted by Daniel Krentzel
Gillen Tener Martin @gillenmartin.bsky.social · 20/08/2026
I wrote some words from a village below a wildfire that are out in the @washingtonmonthly.bsky.social today. washingtonmonthly.com/2026/08/20/h...
011
Reposted by Daniel Krentzel
Joel Lüthi @joelluethi.bsky.social · 14/08/2026
We're excited to announce the 2026 OME-Zarr hackathon 🎉 📅 Date: November 2nd - November 6th 2026 📍 Location: University of Zurich, Zurich, Switzerland 💻 Virtual participation: possible for Monday afternoon, 1:00 PM CET 6:00 PM CET biovisioncenter.notion.site/2026-ome-zar... (1/6 🧵)
2025 OME-Zarr hackathon
11414
Reposted by Daniel Krentzel
Guillaume Jacquemet @guijacquemet.bsky.social · 30/06/2026
Delighted to share our latest preprint "NucleiSky enables cross-scale multimodal registration of microscopy data using nuclei constellations" www.biorxiv.org/content/10.6... Code and App: github.com/CellMigratio...
37436
Reposted by Daniel Krentzel
Aafke Gros @aafkegros.bsky.social · 24/07/2026
Microscopy Nodes 3.1 is out for Blender 5.2! 🎉 This has cool new features for conditional masking, working with even larger data (regional upscaling), and data manipulation! Watch the new youtube tutorials that explain the features: www.youtube.com/playlist?lis...
A FIB-SEM plankton visualization where the top shows only the volumetric data of the chromosomes, and the bottom a normally sliced region of the cell. A fluorescence microscopy dataset where a region has been reloaded to be much higher resolution
26827
Reposted by Daniel Krentzel
Joel Lüthi @joelluethi.bsky.social · 24/07/2026
Do you want to learn more about OME-Zarr viewers, where they stand today and what's being developed at the moment? On August 24th at 4pm CEST, we will hear about 3 different OME-Zarr viewers and user stories for each. www.biovisioncenter.uzh.ch/en/events/Up... (1/5 🧵)
biovisioncenter.uzh.ch
Open Bioimaging Practices Meetup August 2026
11511
Reposted by Daniel Krentzel
Spyros Lytras @spyroslytras.bsky.social · 23/07/2026
Come work on exciting AI for biology projects in Paris!! 🇫🇷 Details and application below, deadline sept 1st!
02727
Reposted by Daniel Krentzel
Estibaliz Gómez de Mariscal, PhD @gomez-mariscal.bsky.social · 24/04/2026
Want to do a PhD in biomedical research and explore the exciting world of AI, microscopy and cell biology?👾🧠🧫🔬 In a research environment of high excellence🌟 and in beautiful Portugal🏄🌸? PhD fellowships are now open at #NIMSB from May-4 til June-5 👉 APPLY lnkd.in/eqkYreFi Reach out if interested!✉️
044
Reposted by Daniel Krentzel
Vivek Mutalik @vivekmutalik.bsky.social · 03/04/2026
📣Huge preprint 🔔 Today we share something our group has been working toward for a long time, led by @lucasmoriniere.bsky.social We asked can we predict which receptor a phage targets from its genome sequence alone? For most phages, we couldn’t. So Lucas set out to do something I had only dreamed of.
Phage receptor prediction from genome sequencing alone. Bacterial receptor (blue) interacting with phage proteins (purple) is shown here
6226120
Reposted by Daniel Krentzel
Washington Monthly @washingtonmonthly.bsky.social · 30/03/2026
Trump’s bid to strip birthright citizenship from millions is a moral, administrative, and legal catastrophe that the Fourteenth Amendment was enacted to prevent. That’s why he’s pursuing it, writes Legal Affairs Editor @garrettepps.bsky.social.
washingtonmonthly.com
Birthright Citizenship: Supreme Court Could Create an Exploitable Noncitizen Class
Trump’s bid to strip birthright citizenship from millions is a moral, administrative, and legal catastrophe.
12317
Reposted by Daniel Krentzel
Levayer Lab @levayerr.bsky.social · 27/03/2026
Happy to share the heroic effort of @gaellel.bsky.social to reduce our segmentation pain through EpiCure (Epithelial Curation), a napari plugin easing the curation of epithelial segmentation developed with several groups of @devstempasteur.bsky.social. www.biorxiv.org/content/10.6... See more 👇
biorxiv.org
14218
Reposted by Daniel Krentzel
Carles Bosch @carlesbosch.bsky.social · 24/03/2026
#paper: Drying tissue samples enhances contrast in X-ray phase contrast imaging while largely preserving ultrastructure. 🧠🔬 doi.org/10.1107/S160... @safekhan.bsky.social @andreas-t-schaefer.bsky.social @crick.ac.uk @embl.org @imperialcollegeldn.bsky.social @esrf.fr @uclnpp.bsky.social
1126
Reposted by Daniel Krentzel
Institut Pasteur | 130 years of biomedical research @pasteur.fr · 23/03/2026
🧬 EMBO Workshop – The complexity of mycobacterial infections: from research to real-world impact 📅 14–18 Sept 2026, Paris Tuberculosis & non-tuberculous mycobacteria at the heart of discussions: evolution, host–pathogen interactions, drug resistance. 🔗 meetings.embo.org/event/26-myc...
Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains one of the deadliest infectious diseases worldwide. Meanwhile, non-tuberculous mycobacteria (NTM) are an increasing global health concern. NTM infections often resemble TB clinically, which can lead to misdiagnosis and inappropriate therapy. Additionally, drug-resistant strains of M. tuberculosis and NTM exacerbate the global antimicrobial resistance crisis. A better understanding of the biology of mycobacterial pathogens and of the mechanisms underlying disease pathogenesis and transmission is critical to reduce their burden worldwide.

This EMBO Workshop is the fourth in a successful series held at the Institut Pasteur in Paris, following the meetings organized in 2012, 2016, and 2022. It will bring together scientists, clinicians, public health experts, and representatives from industry and non-profit organizations, who will share recent insights into mycobacterial infections, from fundamental discoveries to translational research, and their practical implications. The program will cover mycobacterial evolution, physiology, metabolism, host–pathogen interactions, immune responses, mechanisms of drug evasion, therapeutic strategies, vaccines, and diagnostics. Emerging experimental models, computational approaches, and artificial intelligence will also be addressed, showing how recent technologies contribute to advancing mycobacterial research and interventions.
1911
Reposted by Daniel Krentzel
Juan Nunez-Iglesias @jni.codes · 16/03/2026
So… um… I (and Draga Doncila Pop and Eric Perlman and @joshmoore.bsky.social and @kevinyamauchi.bsky.social) started a thing! We are excited and terrified in equal measures, and look forward to working with you all in our amazing scientific imaging community! ❤️‍🔥 Read the linked post for the lowdown.
0156
Daniel Krentzel @danielkrentzel.bsky.social · 15/03/2026
Thank you!! 🙏
011
Reposted by Daniel Krentzel
Caterina Fuster Barceló @cfusterbarcelo.bsky.social · 15/03/2026
Are FMs actually FMing? ⛓️🔬 Thrilled to share that our paper on Foundational Models for mitochondria segmentation in EM is heading to the @iclr-conf.bsky.social 2026 LMRL Workshop! 🚀 Check out the full work here: 📄 ArXiv: arxiv.org/abs/2602.08505 📄 OpenReview: openreview.net/forum?id=Bba...
arxiv.org
Are Vision Foundation Models Foundational for Electron Microscopy Image Segmentation?
Although vision foundation models (VFMs) are increasingly reused for biomedical image analysis, it remains unclear whether the latent representations they provide are general enough to support effecti...
195
Reposted by Daniel Krentzel
David Stansby @dstansby.bsky.social · 13/03/2026
The paper that sums up what I've been working on and running for the last three years was published this week 🎉 www.science.org/doi/10.1126/...
science.org
The Human Organ Atlas
The Human Organ Atlas provides open access to 3D images of human anatomy spanning whole organs to cellular scale resolution.
0198
Reposted by Daniel Krentzel
Joel Lüthi @joelluethi.bsky.social · 10/03/2026
We’re excited to share our latest preprint on Fractal: our approach towards FAIR bioimage analysis at scale with OME-Zarr-native workflows. Fractal defines interoperable tasks on OME-Zarr and provides a platform for TB-scale image analysis. (1 / 10🧵) www.biorxiv.org/content/10.6...
biorxiv.org
23217