Acinetobacter baumannii is a nosocomial Gram-negative pathogen that often displays multidrug resistance. Discovering new antibiotics against A. baumannii has proven challenging through conventional screening approaches. Fortunately, machine learning methods allow for the rapid exploration of chemical space, increasing the probability of discovering new antibacterial molecules. Here we screened ~7,500 molecules for those that inhibited the growth of A. baumannii in vitro. We trained a neural network with this growth inhibition dataset and performed in silico predictions for structurally new molecules with activity against A. baumannii. Through this approach, we discovered abaucin, an antibacterial compound with narrow-spectrum activity against A. baumannii. Further investigations revealed that abaucin perturbs lipoprotein trafficking through a mechanism involving LolE. Moreover, abaucin could control an A. baumannii infection in a mouse wound model. This work highlights the utility of machine learning in antibiotic discovery and describes a promising lead with targeted activity against a challenging Gram-negative pathogen.
This is a preview of subscription content, access via your institution
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 12 print issues and online access
$259.00 per year
only $21.58 per issue
Rent or buy this article
Get just this article for as long as you need it
Prices may be subject to local taxes which are calculated during checkout
GenBank accession numbers for sequencing of abaucin-resistant mutants are BankIt2629921 – OP677864, OP677865, OP677866 and OP677867. GEO accession numbers for RNA sequencing datasets are GSE214305 – GSM6603484, GSM6603485, GSM6603486, GSM6603487, GSM6603488, GSM6603489 and GSM6603490. Source data are provided with this paper.
All custom code used for antibiotic prediction is open source and can be accessed without restriction at https://github.com/chemprop/chemprop. A cloned snapshot used for this paper is available at https://github.com/GaryLiu152/chemprop_abaucin. All commercial software used is described in Methods. Source data are provided with this paper.
Antunes, L. C. S., Visca, P. & Towner, K. J. Acinetobacter baumannii: evolution of a global pathogen. Pathog. Dis. 71, 292–301 (2014).
2020 Antibacterial Agents in Clinical and Preclinical Development: An Overview and Analysis (World Health Organization, 2021); https://www.who.int/publications/i/item/9789240021303
Walsh, C. Where will new antibiotics come from? Nat. Rev. Microbiol. 1, 65–70 (2003).
Tommasi, R., Brown, D. G., Walkup, G. K., Manchester, J. I. & Miller, A. A. ESKAPEing the labyrinth of antibacterial discovery. Nat. Rev. Drug Discov. 14, 529–542 (2015).
Stokes, J. M. et al. A deep learning approach to antibiotic discovery. Cell 180, 688–702 (2020).
Ma, Y. et al. Identification of antimicrobial peptides from the human gut microbiome using deep learning. Nat. Biotechnol. 40, 921–931 (2022).
Lluka, T. & Stokes, J. M. Antibiotic discovery in the artificial intelligence era. Ann. N. Y. Acad. Sci. 1519, 74–93 (2023).
Melander, R. J., Zurawski, D. V. & Melander, C. Narrow-spectrum antibacterial agents. MedChemComm 9, 12–21 (2018).
Theriot, C. M. et al. Antibiotic-induced shifts in the mouse gut microbiome and metabolome increase susceptibility to Clostridium difficile infection. Nat. Commun. 5, 3114 (2014).
Willing, B. P. et al. A pyrosequencing study in twins shows that gastrointestinal microbial profiles vary with inflammatory bowel disease phenotypes. Gastroenterology 139, 1844–1854 (2010).
Turnbaugh, P. J. et al. An obesity-associated gut microbiome with increased capacity for energy harvest. Nature 444, 1027–1031 (2006).
Kelly, J. R. et al. Transferring the blues: depression-associated gut microbiota induces neurobehavioural changes in the rat. J. Psychiatr. Res. 82, 109–118 (2016).
Wu, N. et al. Dysbiosis signature of fecal microbiota in colorectal cancer patients. Microb. Ecol. 66, 462–470 (2013).
Lee, H. S., Plechot, K., Gohil, S. & Le, J. Clostridium difficile: diagnosis and the consequence of over diagnosis. Infect. Dis. Ther. 10, 687–697 (2021).
Corsello, S. M. et al. The Drug Repurposing Hub: a next-generation drug library and information resource. Nat. Med. 23, 405–408 (2017).
Kaplan, E., Greene, N. P., Crow, A. & Koronakis, V. Insights into bacterial lipoprotein trafficking from a structure of LolA bound to the LolC periplasmic domain. Proc. Natl Acad. Sci. USA 115, E7389–E7397 (2018).
Tang, X. et al. Structural basis for bacterial lipoprotein relocation by the transporter LolCDE. Nat. Struct. Mol. Biol. 28, 347–355 (2021).
Yang, K. et al. Analyzing learned molecular representations for property prediction. J. Chem. Inf. Model. 59, 3370–3388 (2019).
Landrum, G. RDKit: a software suite for cheminformatics, computational chemistry, and predictive modeling. https://www.rdkit.org/RDKit_Overview.pdf
Seok, S. J. et al. Blockade of CCL2/CCR2 signalling ameliorates diabetic nephropathy in db/db mice. Nephrol. Dial. Transplant. 28, 1700–1710 (2013).
Cerri, C. et al. The chemokine CCL2 mediates the seizure-enhancing effects of systemic inflammation. J. Neurosci. 36, 3777–3788 (2016).
Chargari, C. et al. Preclinical assessment of JNJ-26854165 (Serdemetan), a novel tryptamine compound with radiosensitizing activity in vitro and in tumor xenografts. Cancer Lett. 312, 209–218 (2011).
Lehman, J. A. et al. Serdemetan antagonizes the Mdm2-HIF1α axis leading to decreased levels of glycolytic enzymes. PLoS ONE 8, e74741 (2013).
Stokes, J. M., Lopatkin, A. J., Lobritz, M. A. & Collins, J. J. Bacterial metabolism and antibiotic efficacy. Cell Metab. 30, 251–259 (2019).
Zheng, E. J., Stokes, J. M. & Collins, J. J. Eradicating bacterial persisters with combinations of strongly and weakly metabolism-dependent antibiotics. Cell Chem. Biol. 27, 1544–1552 (2020).
Francino, M. P. Antibiotics and the human gut microbiome: dysbioses and accumulation of resistances. Front. Microbiol. 6, 1543 (2016).
Smits, W. K., Lyras, D., Lacy, D. B., Wilcox, M. H. & Kuijper, E. J. Clostridium difficile infection. Nat. Rev. Dis. Prim. 2, 16020 (2016).
Sharma, S. et al. Mechanism of LolCDE as a molecular extruder of bacterial triacylated lipoproteins. Nat. Commun. 12, 4687 (2021).
Nicholson, W. L. & Maughan, H. The spectrum of spontaneous rifampin resistance mutations in the rpoB gene of Bacillus subtilis 168 spores differs from that of vegetative cells and resembles that of Mycobacterium tuberculosis. J. Bacteriol. 184, 4936–4940 (2002).
Baek, M. et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science 373, 871–876 (2021).
Buel, G. R. & Walters, K. J. Can AlphaFold2 predict the impact of missense mutations on structure? Nat. Struct. Mol. Biol. 29, 1–2 (2022).
Raivio, T. L., Leblanc, S. K. D. & Price, N. L. The Escherichia coli Cpx envelope stress response regulates genes of diverse function that impact antibiotic resistance and membrane integrity. J. Bacteriol. 195, 2755–2767 (2013).
Guest, R. L., Wang, J., Wong, J. L. & Raivio, T. L. A bacterial stress response regulates respiratory protein complexes to control envelope stress adaptation. J. Bacteriol. 199, e00153-17 (2017).
Delhaye, A., Laloux, G. & Collet, J.-F. The lipoprotein NlpE is a Cpx sensor that serves as a sentinel for protein sorting and folding defects in the Escherichia coli envelope. J. Bacteriol. 201, e00611-18 (2019).
Peters, J. M. et al. A comprehensive, CRISPR-based functional analysis of essential genes in bacteria. Cell 165, 1493–1506 (2016).
Pathania, R. et al. Chemical genomics in Escherichia coli identifies an inhibitor of bacterial lipoprotein targeting. Nat. Chem. Biol. 5, 849–856 (2009).
McLeod, S. M. et al. Small-molecule inhibitors of Gram-negative lipoprotein trafficking discovered by phenotypic screening. J. Bacteriol. 197, 1075–1082 (2015).
Manchanda, V., Sanchaita, S. & Singh, N. Multidrug resistant acinetobacter. J. Glob. Infect. Dis. 2, 291–304 (2010).
Davis, K. A., Moran, K. A., McAllister, C. K. & Gray, P. J. Multidrug-resistant Acinetobacter extremity infections in soldiers. Emerg. Infect. Dis. 11, 1218–1224 (2005).
Jin, W. et al. Deep learning identifies synergistic drug combinations for treating COVID-19. Proc. Natl Acad. Sci. USA 118, e2105070118 (2021).
Preuer, K. et al. DeepSynergy: predicting anti-cancer drug synergy with Deep Learning. Bioinformatics 34, 1538–1546 (2018).
Jin, W., Barzilay, R. & Jaakkola, T. Junction tree variational autoencoder for molecular graph generation. In Proc. 35th International Conference on Machine Learning (eds Dy, J. & Krause, A.) 2323–2332 (PMLR, 2018).
Pedregosa, F. et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011).
Deatherage, D. E. & Barrick, J. E. Identification of mutations in laboratory-evolved microbes from next-generation sequencing data using breseq. Methods Mol. Biol. 1151, 165–188 (2014).
Bray, N. L., Pimentel, H., Melsted, P. & Pachter, L. Near-optimal probabilistic RNA-seq quantification. Nat. Biotechnol. 34, 525–527 (2016).
Ihaka, R. & Gentleman, R. R: a language for data analysis and graphics. J. Comput. Graph. Stat. 5, 299–314 (1996).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Karp, P. D. Pathway databases: a case study in computational symbolic theories. Science 293, 2040–2044 (2001).
Keseler, I. M. et al. EcoCyc: fusing model organism databases with systems biology. Nucleic Acids Res. 41, D605–D612 (2013).
Karp, P. D. et al. Pathway tools version 19.0 update: software for pathway/genome informatics and systems biology. Brief. Bioinform. 17, 877–890 (2016).
Calvo-Villamañán, A. et al. On-target activity predictions enable improved CRISPR–dCas9 screens in bacteria. Nucleic Acids Res. 48, e64 (2020).
Depardieu, F. & Bikard, D. Gene silencing with CRISPRi in bacteria and optimization of dCas9 expression levels. Methods 172, 61–75 (2020).
UniProt Consortium. UniProt: the universal protein knowledgebase in 2021. Nucleic Acids Res. 49, D480–D489 (2021).
Altschul, S. F. et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25, 3389–3402 (1997).
Altschul, S. F. et al. Protein database searches using compositionally adjusted substitution matrices. FEBS J. 272, 5101–5109 (2005).
Goddard, T. D. et al. UCSF ChimeraX: meeting modern challenges in visualization and analysis. Protein Sci. 27, 14–25 (2018).
Pettersen, E. F. et al. UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82 (2021).
We thank S. French from McMaster University for technical assistance with fluorescence microscopy experiments. This work was supported by the David Braley Centre for Antibiotic Discovery (to J.M.S.); the Weston Family Foundation (POP and Catalyst to J.M.S.); the Audacious Project (to J.J.C. and J.M.S.); the C3.ai Digital Transformation Institute (to R.B.); the Abdul Latif Jameel Clinic for Machine Learning in Health (to R.B.); the DTRA Discovery of Medical Countermeasures Against New and Emerging (DOMANE) threats program (to R.B.); the DARPA Accelerated Molecular Discovery program (to R.B.); the Canadian Institutes of Health Research (FRN-156361 to B.K.C.); Genome Canada GAPP (OGI-146 to M.G.S.); the Canadian Institutes of Health Research (FRN-148713 to M.G.S.); the Faculty of Health Sciences of McMaster University (to J.M.); the Boris Family (to J.M.); a Marshall Scholarship (to K.S.); and the DOE BER (DE-FG02-02ER63445 to A.C-P.).
J.M.S. is cofounder and scientific director of Phare Bio. J.J.C. is cofounder and scientific advisory board chair of Phare Bio. J.J.C. is cofounder and scientific advisory board chair of Enbiotix. The other authors declare no competing interests.
Peer review information
Nature Chemical Biology thanks Jean Francois Collet and the other, anonymous reviewers for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended Data Fig. 1 Model training data and prediction.
(a) Replicate plot showing primary screening data of 7,684 small molecules for those that inhibited the growth of A. baumannii ATCC 17978 in LB medium at 50 µM. (b) Rank-ordered growth inhibition data of the prioritized 240 molecules from our prediction set that were selected for empirical validation (top); rank-ordered growth inhibition data of the 240 predicted molecules with the lowest prediction score (middle); rank-ordered growth inhibition data of the 240 predicted molecules with the highest prediction score that were not found in the training dataset (bottom). Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted. Dashed horizontal line represents the stringent hit cut-off of >80% growth inhibition at 50 µM. (c) Growth inhibition of A. baumannii by abaucin (blue) and serdemetan (red) in LB medium. Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted. The structure of serdemetan is shown. (d) Growth kinetics of A. baumannii cells after treatment with abaucin at varying concentrations for 6 hours. Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted.
Extended Data Fig. 2 Antibacterial activity of abaucin against human commensal species.
(a) Growth inhibition of A. baumannii ATCC 17978 by ampicillin (blue) and ciprofloxacin (red) in LB medium. Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted. (b) Growth inhibition of B. breve by abaucin. Experiments were conducted in biological duplicate. (c) Growth inhibition of B. longum by abaucin. Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted. (d) Non-validated (see Fig. 2e) growth inhibition of E. lenta by abaucin. Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted.
Extended Data Fig. 3 Abaucin mechanism of action.
(a–h) Growth inhibition of wildtype A. baumannii (WT) and the four independent abaucin-resistant mutants by a collection of diverse antibiotics. From left to right for each plot, the mutants are: A362T variant 1, Y394F, intergenic, and A362T variant 2. Experiments were conducted in biological duplicate. Note that the abaucin-resistant mutants do not display cross-resistance to other antibiotics. (i) Structural prediction of wildtype A. baumannii LolE using RoseTTAFold (bottom), with the structural error estimate of each amino acid (top). Position 362 is highlighted orange and resides in a disordered region of the protein. (j) same as (i), except with the Y362T abaucin-resistant mutant of LolE. (k) RNA sequencing of wildtype A. baumannii treated with 5x MIC abaucin for 4.5 hr (top) or 6 hr (bottom). Data are the mean of biological duplicates. Transcript abundance is normalized to no-drug control cultures grown in identical conditions. Vertical black lines show statistical significance cut-off values. Note the highly significant downregulation of genes involved in the electron transport chain and transmembrane ion transport. (l) Growth inhibition of A. baumannii harboring an empty CRISPRi vector (red), or three distinct sgRNAs targeting lolE (blue, teal, and green). All strains were grown in LB medium without induction. Experiments were conducted in biological duplicate. Individual replicates with means connected are plotted. (m) qPCR quantifying the expression of lolE relative to the housekeeping gene gltA (left) and gyrB (right) in all four abaucin resistant mutants, normalized to wildtype A. baumannii. Experiments were conducted in biological duplicate with technical triplicates. Bar height represents mean expression.
Supplementary Tables 1–7 and Note.
Supplementary Data 1: Growth inhibition data against A. baumannii for model training. Supplementary Data 2: Model prediction scores of compounds in the Drug Repurposing Hub. Supplementary Data 3: Experimental validation of (prioritized/poorest/top 240) predictions from the Drug Repurposing Hub. Supplementary Data 4: GO enrichment for up- and down-regulated transcripts in A. baumannii treated with 5x MIC abaucin.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Liu, G., Catacutan, D.B., Rathod, K. et al. Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii. Nat Chem Biol (2023). https://doi.org/10.1038/s41589-023-01349-8