Discovery of a structural class of antibiotics with explainable deep learning.

The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis1-9. Deep learning approaches have aided in exploring chemical spaces1,10-15; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of 39,312 compounds and applied ensembles of graph neural networks to predict antibiotic activity and cytotoxicity for 12,076,365 compounds. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against Staphylococcus aureus were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant S. aureus (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.

Global diversity of enterococci and description of 18 previously unknown species.

Enterococci are gut microbes of most land animals. Likely appearing first in the guts of arthropods as they moved onto land, they diversified over hundreds of millions of years adapting to evolving hosts and host diets. Over 60 enterococcal species are now known. Two species, Enterococcus faecalis and Enterococcus faecium, are common constituents of the human microbiome. They are also now leading causes of multidrug-resistant hospital-associated infection. The basis for host association of enterococcal species is unknown. To begin identifying traits that drive host association, we collected 886 enterococcal strains from widely diverse hosts, ecologies, and geographies. This identified 18 previously undescribed species expanding genus diversity by >25%. These species harbor diverse genes including toxins and systems for detoxification and resource acquisition. Enterococcus faecalis and E. faecium were isolated from diverse hosts highlighting their generalist properties. Most other species showed a more restricted distribution indicative of specialized host association. The expanded species diversity permitted the Enterococcus genus phylogeny to be viewed with unprecedented resolution, allowing features to be identified that distinguish its four deeply rooted clades, and the entry of genes associated with range expansion such as B-vitamin biosynthesis and flagellar motility to be mapped to the phylogeny. This work provides an unprecedentedly broad and deep view of the genus Enterococcus, including insights into its evolution, potential new threats to human health, and where substantial additional enterococcal diversity is likely to be found.

SAFPred: synteny-aware gene function prediction for bacteria using protein embeddings.

MOTIVATION: Today, we know the function of only a small fraction of the protein sequences predicted from genomic data. This problem is even more salient for bacteria, which represent some of the most phylogenetically and metabolically diverse taxa on Earth. This low rate of bacterial gene annotation is compounded by the fact that most function prediction algorithms have focused on eukaryotes, and conventional annotation approaches rely on the presence of similar sequences in existing databases. However, often there are no such sequences for novel bacterial proteins. Thus, we need improved gene function prediction methods tailored for bacteria. Recently, transformer-based language models-adopted from the natural language processing field-have been used to obtain new representations of proteins, to replace amino acid sequences. These representations, referred to as protein embeddings, have shown promise for improving annotation of eukaryotes, but there have been only limited applications on bacterial genomes. RESULTS: To predict gene functions in bacteria, we developed SAFPred, a novel synteny-aware gene function prediction tool based on protein embeddings from state-of-the-art protein language models. SAFpred also leverages the unique operon structure of bacteria through conserved synteny. SAFPred outperformed both conventional sequence-based annotation methods and state-of-the-art methods on multiple bacterial species, including for distant homolog detection, where the sequence similarity to the proteins in the training set was as low as 40%. Using SAFPred to identify gene functions across diverse enterococci, of which some species are major clinical threats, we identified 11 previously unrecognized putative novel toxins, with potential significance to human and animal health. AVAILABILITY AND IMPLEMENTATION: https://github.com/AbeelLab/safpred.

Bacterial strain sharing between humans, animals, and the environment among urban households.

Identifying bacterial transmission pathways is crucial to inform strategies aimed at curbing the spread of pathogenic and antibiotic-resistant bacteria, especially in rapidly urbanizing low- and middle-income countries. In this study, we assessed bacterial strain-sharing and dissemination of antibiotic resistance across humans, domesticated poultry, canines, household soil, and drinking water in urban informal settlements in Nairobi, Kenya. We collected 321 samples from 50 households and performed Pooling Isolated Colonies-seq (PIC-seq) by sequencing pools of up to five Escherichia coli colonies per sample to capture strain diversity, strain-sharing patterns, and overlap of antibiotic-resistant genes (ARGs). Bacterial strains isolated from the household environment carried clinically relevant ARGs, reinforcing the role of the environment in antibiotic resistance dissemination. Strain-sharing rates and resistome similarities across sample types were strongly correlated within households, suggesting clonal spread of bacteria is a main driver of dissemination of ARGs in the domestic urban environment. Within households, E. coli strain-sharing was rare between humans and animals but more frequent between humans and drinking water. E. coli contamination in stored drinking water was also associated with higher strain-sharing between humans in the same household. Our study demonstrates that contaminated drinking water facilitates human to human strain sharing and water treatment can disrupt transmission.

Dynamics of microbiome and resistome in a poultry burger processing line.

Traditionally, surveillance programs for food products and food processing environments have focused on targeted pathogens and resistance genes. Recent advances in high throughput sequencing allow for more comprehensive and untargeted monitoring. This study assessed the microbiome and resistome in a poultry burger processing line using culturing techniques and whole metagenomic sequencing (WMS). Samples included meat, burgers, and expired burgers, and different work surfaces. Microbiome analysis revealed spoilage microorganisms as the main microbiota, with substantial shifts observed during the shelf-life period. Core microbiota of meat and burgers included Pseudomonas spp., Psychrobacter spp., Shewanella spp. and Brochothrix spp., while expired burgers were dominated by Latilactobacillus spp. and Leuconostoc spp. Cleaning and disinfection (C&D) procedures altered the microbial composition of work surfaces, which still harbored Hafnia spp. and Acinetobacter spp. after C&D. Resistome analysis showed a low overall abundance of resistance genes, suggesting that effective interventions during processing may mitigate their transmission. However, biocide resistance genes were frequently found, indicating potential biofilm formation or inefficient C&D protocols. This study demonstrates the utility of combining culturing techniques and WMS for comprehensive of the microbiome and resistome characterization in food processing lines.

Persistent Salmonella infections in humans are associated with mutations in the BarA/SirA regulatory pathway.

Several bacterial pathogens, including Salmonella enterica, can cause persistent infections in humans by mechanisms that are poorly understood. By comparing genomes of isolates longitudinally collected from 256 prolonged salmonellosis patients, we identified repeated mutations in global regulators, including the barA/sirA two-component regulatory system, across multiple patients and Salmonella serovars. Comparative RNA-seq analysis revealed that distinct mutations in barA/sirA led to diminished expression of Salmonella pathogenicity islands 1 and 4 genes, which are required for Salmonella invasion and enteritis. Moreover, barA/sirA mutants were attenuated in an acute salmonellosis mouse model and induced weaker transcription of host immune responses. In contrast, in a persistent infection mouse model, these mutants exhibited long-term colonization and prolonged shedding. Taken together, these findings suggest that selection of mutations in global virulence regulators facilitates persistent Salmonella infection in humans, by attenuating Salmonella virulence and inducing a weaker host inflammatory response.

Distinct Escherichia coli transcriptional profiles in the guts of recurrent UTI sufferers revealed by pangenome hybrid selection.

Low-abundance members of microbial communities are difficult to study in their native habitats. This includes Escherichia coli, a minor, but common inhabitant of the gastrointestinal tract and opportunistic pathogen, including of the urinary tract, where it is the primary pathogen. While multi-omic analyses have detailed critical interactions between uropathogenic Escherichia coli (UPEC) and the bladder that mediate UTI outcome, comparatively little is known about UPEC in its pre-infection reservoir, partly due to its low abundance there (<1% relative abundance). To accurately and sensitively explore the genomes and transcriptomes of diverse E. coli in gastrointestinal communities, we developed E. coli PanSelect which uses a set of probes designed to specifically recognize and capture E. coli's broad pangenome from sequencing libraries. We demonstrated the ability of E. coli PanSelect to enrich, by orders of magnitude, sequencing data from diverse E. coli using a mock community and a set of human stool samples collected as part of a cohort study investigating drivers of recurrent urinary tract infections (rUTI). Comparisons of genomes and transcriptomes between E. coli residing in the gastrointestinal tracts of women with and without a history of rUTI suggest that rUTI gut E. coli are responding to increased levels of oxygen and nitrate, suggestive of mucosal inflammation, which may have implications for recurrent disease. E. coli PanSelect is well suited for investigations of native in vivo biology of E. coli in other environments where it is at low relative abundance, and the framework described here has broad applicability to other highly diverse, low abundance organisms.