2025 Hyper Recent •CC0 1.0 Universal

This work is dedicated to the public domain. No rights reserved.

Access Preprint From Server
June 7th, 2025
Version: 3
Institut Pasteur
microbiology
biorxiv

Deep learning recognises antibiotic modes of action from brightfield images

Krentzel, D.Open in Google Scholar•Kho, K.Open in Google Scholar•Petit, J.Open in Google Scholar•Mahtal, N.Open in Google Scholar•Chiaravalli, J.Open in Google Scholar•Shorte, S. L.Open in Google Scholar•Wehenkel, A. M.Open in Google Scholar•Boneca, I. G.Open in Google Scholar•Zimmer, C.Open in Google Scholar

Antimicrobial resistance is a growing public health threat predicted to cause up to 10 million deaths a year by 2050. To circumvent existing bacterial resistance mechanisms, discovering antibiotics with novel modes of action (MoAs) is crucial. While growth inhibition assays can robustly identify antibiotic molecules, they miss promising compounds with subinhibitory phenotypes and do not inform on drug MoA. Microscopy-based cytological profiling of drug-treated bacteria with hand-crafted image descriptors or more recently deep learning (DL) provides complementary information on the MoA. However, current approaches are limited by the need for fluorescent labelling and drug exposure at inhibitory concentrations. It also remains unclear if cytological profiling enables the detection of drugs with novel MoAs. Here, we demonstrate an approach based on supervised DL to identify antibiotic drug MoA from microscopy images without fluorescent labelling. We train a convolutional neural network to predict treatment conditions from brightfield images of Escherichia coli exposed to reference drugs covering multiple MoAs. Our method can detect drug exposure at subinhibitory concentrations and distinguishes individual drug treatments with high accuracy (86.2%). The learned representations implicitly capture MoA-specific phenotypes, enabling perfect MoA recognition (100%) without retraining using a model trained on only 644 images. Our approach can identify the MoA of previously unseen drugs with good accuracy (77.8{+/-}3.3%), as long as the MoA is represented by at least one of the training compounds. Finally, we show that our approach can detect if a drug has a novel MoA with an area under the curve above 0.75 for five out of six MoAs, facilitating microscopy-based identification of novel classes of antibiotics. Our methods and results pave the way towards an automated pipeline for antibiotic drug discovery based on imaging and DL.

Similar Papers

biorxiv
Sat Jun 07 2025
TICKHUNTER: A Targeted Hybridization-Capture Sequencing Approach for the Detection and Characterization of Tick-borne Pathogens and Blood Meals
Ticks are vectors of many debilitating pathogens that are transmitted during blood meal acquisition. As both vector and pathogen species continue to emerge, there is a critical need for improved pathogen diagnostic and host remnant identification methods to develop interventions. Current pathogen detection methods are limited in their ability to subtype and detect emerging variants and non-vector ...
Wang, X.
•
Brown, J. J.
•
Lilly, M. V.
•
Plimpton, L.
...•
Goodman, L. B.
biorxiv
Sat Jun 07 2025
Analysis of Essential Genes in Clostridioides difficile by CRISPRi and Tn-seq
Essential genes are interesting in their own right and as potential antibiotic targets. To date, only one report has identified essential genes on a genome-wide scale in Clostridioides difficile, a problematic pathogen for which treatment options are limited. That foundational study used large-scale transposon mutagenesis to identify 404 protein-encoding genes as likely to be essential for vegetat...
Alberts, M.
•
Kurtz, M.
•
Müh, U.
•
Bernardi, J.
...•
Weiss, D. S.
biorxiv
Sat Jun 07 2025
Heme-Mediated Selection of Encapsulated Streptococcus pneumoniae in the Lungs by Oxidative Stress
Streptococcus pneumoniae (the pneumococcus) causes cytotoxicity and encapsulates within the lung parenchyma, leading to pneumococcal pneumonia. However, the underlying mechanisms remain unclear and likely involve multiple bacterial and host factors. We investigated the selection process of encapsulated pneumococci, a critical factor in lung damage during pneumococcal pneumonia. Our study revealed ...
Alibayov, B.
•
Jop Vidal, A. G.
•
Murin, L.
•
Scasny, A.
...•
Vidal, J. E.
biorxiv
Fri Jun 06 2025
Metabolic controls on the carbon isotope fractionations of bacterial fermentation
Microbial fermentation facilitates the initial breakdown of organic matter into small molecules, and is thought to be the rate-limiting step for mineralization under anoxic conditions. Fermentation is under studied in modern and ancient biogeochemistry due to a lack of environmental biomarkers that would constrain its activity. It has long been assumed that fermentation, like respiration, does not...
Mueller, E.
•
Heuer, V. B.
•
Leadbetter, J. R.
•
Hinrichs, K.-U.
•
Sessions, A. L.
biorxiv
Fri Jun 06 2025
The study of the Drosophila innubila Nudivirus in cells and flies
Species of the fruit fly, Drosophila, are essential models for investigating host/virus interactions. Research aimed at understanding how hosts mount immune defenses against viruses and how viruses evade those immune defenses relies on host/virus pairs that have evolved together overtime, thus necessitating a need for a DNA virus model system for Drosophila with a virus that naturally infects the ...
Schedl, M. E.
•
Mulkey, K. M.
•
Unckless, R. L.
biorxiv
Fri Jun 06 2025
Rationally Designed, ACE2 Mimetic Binder to the SARS Cov-2 Associated Spike Protein for COVID-19 Therapeutics and Beyond
We have developed a 23 residue, optimized helical peptide biomimetic from the SARS-CoV-2 Spike protein binding partner ACE2 that demonstrated a Therapeutic Index (TI) of >~20 in authentic viral cell challenge assays, including WT and Omicron strains. The therapeutic is an optimized \"peptide decoy\" based on the viruss human cell target ACE2 and, as such, may have more general applicability across...
Peters, M. H.
•
Pei, X. Y.
biorxiv
Fri Jun 06 2025
Unipolar polysaccharide-mediated attachment of the N2O-reducing bacterium Bradyrhizobium ottawaense SG09 to plant roots.
Agricultural soils are an important source of nitrous oxide (N2O), which has greenhouse and ozone-depleting effects. Bradyrhizobium ottawaense SG09 is a nitrogen-fixing rhizobium with high N2O-reducing activity. Rhizobia form symbiotic nodules in leguminous plants. The initial physical attachment of bacteria to plant roots is a critical step in the establishment of symbiotic interactions. In this ...
Takeguchi, Y.
•
Shibuya, R.
•
Kondo, M.
•
Betsuyaku, E.
...•
Betsuyaku, S.
biorxiv
Fri Jun 06 2025
Phage intervention improves colitis and response to corticosteroids by attenuating virulence of Crohn's disease-associated bacteria
Adherent-invasive Escherichia coli (AIEC) exhibit proinflammatory properties and have been implicated in the pathogenesis of Crohns disease (CD), a form of inflammatory bowel disease (IBD). Antibiotic use in CD lacks specificity and may worsen microbiome disruption, prompting interest in bacteriophages (phages) for targeted microbiome editing. Here, we identified HER259, a phage active against the...
Jackson, K.
•
Galipeau, H.
•
Hann, A.
•
Constante, M.
...•
Verdu, E. F.
biorxiv
Fri Jun 06 2025
Establishment of an antimetabolite-based transformation system for the wood-decaying basidiomycete Phanerochaete chrysosporium
The model wood-decaying basidiomycete Phanerochaete chrysosporium has been extensively studied to elucidate the molecular mechanisms of wood decomposition. However, genetic studies have been limited by the lack of adequate genetic tools. Here, we established an antimetabolite-based transformation system, originally developed for ascomycetes, for use in P. chrysosporium. The transformation system u...
Masumoto, K.
•
Banko, P.
•
Yamamoto, A.
•
Miwa, K.
•
Hori, C.