Assessing computational predictions of antimicrobial resistance phenotypes from microbial genomes.

The advent of rapid whole-genome sequencing has created new opportunities for computational prediction of antimicrobial resistance (AMR) phenotypes from genomic data. Both rule-based and machine learning (ML) approaches have been explored for this task, but systematic benchmarking is still needed. Here, we evaluated four state-of-the-art ML methods (Kover, PhenotypeSeeker, Seq2Geno2Pheno and Aytan-Aktug), an ML baseline and the rule-based ResFinder by training and testing each of them across 78 species-antibiotic datasets, using a rigorous benchmarking workflow that integrates three evaluation approaches, each paired with three distinct sample splitting methods. Our analysis revealed considerable variation in the performance across techniques and datasets. Whereas ML methods generally excelled for closely related strains, ResFinder excelled for handling divergent genomes. Overall, Kover most frequently ranked top among the ML approaches, followed by PhenotypeSeeker and Seq2Geno2Pheno. AMR phenotypes for antibiotic classes such as macrolides and sulfonamides were predicted with the highest accuracies. The quality of predictions varied substantially across species-antibiotic combinations, particularly for beta-lactams; across species, resistance phenotyping of the beta-lactams compound, aztreonam, amoxicillin/clavulanic acid, cefoxitin, ceftazidime and piperacillin/tazobactam, alongside tetracyclines demonstrated more variable performance than the other benchmarked antibiotics. By organism, Campylobacter jejuni and Enterococcus faecium phenotypes were more robustly predicted than those of Escherichia coli, Staphylococcus aureus, Salmonella enterica, Neisseria gonorrhoeae, Klebsiella pneumoniae, Pseudomonas aeruginosa, Acinetobacter baumannii, Streptococcus pneumoniae and Mycobacterium tuberculosis. In addition, our study provides software recommendations for each species-antibiotic combination. It furthermore highlights the need for optimization for robust clinical applications, particularly for strains that diverge substantially from those used for training.

Optimized model architectures for deep learning on genomic data.

The success of deep learning in various applications depends on task-specific architecture design choices, including the types, hyperparameters, and number of layers. In computational biology, there is no consensus on the optimal architecture design, and decisions are often made using insights from more well-established fields such as computer vision. These may not consider the domain-specific characteristics of genome sequences, potentially limiting performance. Here, we present GenomeNet-Architect, a neural architecture design framework that automatically optimizes deep learning models for genome sequence data. It optimizes the overall layout of the architecture, with a search space specifically designed for genomics. Additionally, it optimizes hyperparameters of individual layers and the model training procedure. On a viral classification task, GenomeNet-Architect reduced the read-level misclassification rate by 19%, with 67% faster inference and 83% fewer parameters, and achieved similar contig-level accuracy with ~100 times fewer parameters compared to the best-performing deep learning baselines.

A self-supervised deep learning method for data-efficient training in genomics.

Deep learning in bioinformatics is often limited to problems where extensive amounts of labeled data are available for supervised classification. By exploiting unlabeled data, self-supervised learning techniques can improve the performance of machine learning models in the presence of limited labeled data. Although many self-supervised learning methods have been suggested before, they have failed to exploit the unique characteristics of genomic data. Therefore, we introduce Self-GenomeNet, a self-supervised learning technique that is custom-tailored for genomic data. Self-GenomeNet leverages reverse-complement sequences and effectively learns short- and long-term dependencies by predicting targets of different lengths. Self-GenomeNet performs better than other self-supervised methods in data-scarce genomic tasks and outperforms standard supervised training with ~10 times fewer labeled training data. Furthermore, the learned representations generalize well to new datasets and tasks. These findings suggest that Self-GenomeNet is well suited for large-scale, unlabeled genomic datasets and could substantially improve the performance of genomic models.

Pulsed antibiotic treatments of gnotobiotic mice manifest in complex bacterial community dynamics and resistance effects.

Bacteria can evolve to withstand a wide range of antibiotics (ABs) by using various resistance mechanisms. How ABs affect the ecology of the gut microbiome is still poorly understood. We investigated strain-specific responses and evolution during repeated AB perturbations by three clinically relevant ABs, using gnotobiotic mice colonized with a synthetic bacterial community (oligo-mouse-microbiota). Over 80 days, we observed resilience effects at the strain and community levels, and we found that they were correlated with modulations of the estimated growth rate and levels of prophage induction as determined from metagenomics data. Moreover, we tracked mutational changes in the bacterial populations, and this uncovered clonal expansion and contraction of haplotypes and selection of putative AB resistance-conferring SNPs. We functionally verified these mutations via reisolation of clones with increased minimum inhibitory concentration (MIC) of ciprofloxacin and tetracycline from evolved communities. This demonstrates that host-associated microbial communities employ various mechanisms to respond to selective pressures that maintain community stability.

Strain Identification and Quantitative Analysis in Microbial Communities.

Microbiology has long studied the ways in which subtle genetic differences between closely related microbial strains can have profound impacts on their phenotypes and those of their surrounding environments and communities. Despite the growth in high-throughput microbial community profiling, however, such strain-level differences remain challenging to detect. Once detected, few quantitative approaches have been well-validated for associating strain variants from microbial communities with phenotypes of interest, such as medication usage, treatment efficacy, host environment, or health. First, the term "strain" itself is not used consistently when defining a highly-resolved taxonomic or genomic unit from within a microbial community. Second, computational methods for identifying such strains directly from shotgun metagenomics are difficult, with several possible reference- and assembly-based approaches available, each with different sensitivity/specificity tradeoffs. Finally, statistical challenges exist in using any of the resulting strain profiles for downstream analyses, which can include strain tracking, phylogenetic analysis, or genetic association studies. We provide an in depth discussion of recently available computational tools that can be applied for this task, as well as statistical models and gaps in performing and interpreting any of these three main types of studies using strain-resolved shotgun metagenomic profiling of microbial communities.

Rapid and accurate identification of ribosomal RNA sequences via deep learning.

Advances in transcriptomic and translatomic techniques enable in-depth studies of RNA activity profiles and RNA-based regulatory mechanisms. Ribosomal RNA (rRNA) sequences are highly abundant among cellular RNA, but if the target sequences do not include polyadenylation, these cannot be easily removed in library preparation, requiring their post-hoc removal with computational techniques to accelerate and improve downstream analyses. Here, we describe RiboDetector, a novel software based on a Bi-directional Long Short-Term Memory (BiLSTM) neural network, which rapidly and accurately identifies rRNA reads from transcriptomic, metagenomic, metatranscriptomic, noncoding RNA, and ribosome profiling sequence data. Compared with state-of-the-art approaches, RiboDetector produced at least six times fewer misclassifications on the benchmark datasets. Importantly, the few false positives of RiboDetector were not enriched in certain Gene Ontology (GO) terms, suggesting a low bias for downstream functional profiling. RiboDetector also demonstrated a remarkable generalizability for detecting novel rRNA sequences that are divergent from the training data with sequence identities of <90%. On a personal computer, RiboDetector processed 40M reads in less than 6 min, which was ∼50 times faster in GPU mode and ∼15 times in CPU mode than other methods. RiboDetector is available under a GPL v3.0 license at https://github.com/hzi-bifo/RiboDetector.

In vitro interaction network of a synthetic gut bacterial community.

A key challenge in microbiome research is to predict the functionality of microbial communities based on community membership and (meta)-genomic data. As central microbiota functions are determined by bacterial community networks, it is important to gain insight into the principles that govern bacteria-bacteria interactions. Here, we focused on the growth and metabolic interactions of the Oligo-Mouse-Microbiota (OMM12) synthetic bacterial community, which is increasingly used as a model system in gut microbiome research. Using a bottom-up approach, we uncovered the directionality of strain-strain interactions in mono- and pairwise co-culture experiments as well as in community batch culture. Metabolic network reconstruction in combination with metabolomics analysis of bacterial culture supernatants provided insights into the metabolic potential and activity of the individual community members. Thereby, we could show that the OMM12 interaction network is shaped by both exploitative and interference competition in vitro in nutrient-rich culture media and demonstrate how community structure can be shifted by changing the nutritional environment. In particular, Enterococcus faecalis KB1 was identified as an important driver of community composition by affecting the abundance of several other consortium members in vitro. As a result, this study gives fundamental insight into key drivers and mechanistic basis of the OMM12 interaction network in vitro, which serves as a knowledge base for future mechanistic in vivo studies.

Evaluating assembly and variant calling software for strain-resolved analysis of large DNA viruses.

Infection with human cytomegalovirus (HCMV) can cause severe complications in immunocompromised individuals and congenitally infected children. Characterizing heterogeneous viral populations and their evolution by high-throughput sequencing of clinical specimens requires the accurate assembly of individual strains or sequence variants and suitable variant calling methods. However, the performance of most methods has not been assessed for populations composed of low divergent viral strains with large genomes, such as HCMV. In an extensive benchmarking study, we evaluated 15 assemblers and 6 variant callers on 10 lab-generated benchmark data sets created with two different library preparation protocols, to identify best practices and challenges for analyzing such data. Most assemblers, especially metaSPAdes and IVA, performed well across a range of metrics in recovering abundant strains. However, only one, Savage, recovered low abundant strains and in a highly fragmented manner. Two variant callers, LoFreq and VarScan2, excelled across all strain abundances. Both shared a large fraction of false positive variant calls, which were strongly enriched in T to G changes in a 'G.G' context. The magnitude of this context-dependent systematic error is linked to the experimental protocol. We provide all benchmarking data, results and the entire benchmarking workflow named QuasiModo, Quasispecies Metric determination on omics, under the GNU General Public License v3.0 (https://github.com/hzi-bifo/Quasimodo), to enable full reproducibility and further benchmarking on these and other data.

Identification of Natural CRISPR Systems and Targets in the Human Microbiome.

Many bacteria resist invasive DNA by incorporating sequences into CRISPR loci, which enable sequence-specific degradation. CRISPR systems have been well studied from isolate genomes, but culture-independent metagenomics provide a new window into their diversity. We profiled CRISPR loci and cas genes in the body-wide human microbiome using 2,355 metagenomes, yielding functional and taxonomic profiles for 2.9 million spacers by aligning the spacer content to each sample's metagenome and corresponding gene families. Spacer and repeat profiles agree qualitatively with those from isolate genomes but expand their diversity by approximately 13-fold, with the highest spacer load present in the oral microbiome. The taxonomy of spacer sequences parallels that of their source community, with functional targets enriched for viral elements. When coupled with cas gene systems, CRISPR-Cas subtypes are highly site and taxon specific. Our analysis provides a comprehensive collection of natural CRISPR-cas loci and targets in the human microbiome.

Evolutionary Stabilization of Cooperative Toxin Production through a Bacterium-Plasmid-Phage Interplay.

Colicins are toxins produced and released by Enterobacteriaceae to kill competitors in the gut. While group A colicins employ a division of labor strategy to liberate the toxin into the environment via colicin-specific lysis, group B colicin systems lack cognate lysis genes. In Salmonella enterica serovar Typhimurium (S. Tm), the group B colicin Ib (ColIb) is released by temperate phage-mediated bacteriolysis. Phage-mediated ColIb release promotes S. Tm fitness against competing Escherichia coli It remained unclear how prophage-mediated lysis is realized in a clonal population of ColIb producers and if prophages contribute to evolutionary stability of toxin release in S. Tm. Here, we show that prophage-mediated lysis occurs in an S. Tm subpopulation only, thereby introducing phenotypic heterogeneity to the system. We established a mathematical model to study the dynamic interplay of S. Tm, ColIb, and a temperate phage in the presence of a competing species. Using this model, we studied long-term evolution of phage lysis rates in a fluctuating infection scenario. This revealed that phage lysis evolves as bet-hedging strategy that maximizes phage spread, regardless of whether colicin is present or not. We conclude that the ColIb system, lacking its own lysis gene, is making use of the evolutionary stable phage strategy to be released. Prophage lysis genes are highly prevalent in nontyphoidal Salmonella genomes. This suggests that the release of ColIb by temperate phages is widespread. In conclusion, our findings shed new light on the evolution and ecology of group B colicin systems.IMPORTANCE Bacteria are excellent model organisms to study mechanisms of social evolution. The production of public goods, e.g., toxin release by cell lysis in clonal bacterial populations, is a frequently studied example of cooperative behavior. Here, we analyze evolutionary stabilization of toxin release by the enteric pathogen Salmonella The release of colicin Ib (ColIb), which is used by Salmonella to gain an edge against competing microbiota following infection, is coupled to bacterial lysis mediated by temperate phages. Here, we show that phage-dependent lysis and subsequent release of colicin and phage particles occurs only in part of the ColIb-expressing Salmonella population. This phenotypic heterogeneity in lysis, which represents an essential step in the temperate phage life cycle, has evolved as a bet-hedging strategy under fluctuating environments such as the gastrointestinal tract. Our findings suggest that prophages can thereby evolutionarily stabilize costly toxin release in bacterial populations.

Publisher Correction: Genomic variation and strain-specific functional adaptation in the human gut microbiome during early life.

In the version of this Article originally published, in the first sentence of the second paragraph of the Discussion section, the word "operingrationally" should have read "operationally". This has now been amended in all versions of the Article.

DiTaxa: nucleotide-pair encoding of 16S rRNA for host phenotype and biomarker detection.

SUMMARY: Identifying distinctive taxa for micro-biome-related diseases is considered key to the establishment of diagnosis and therapy options in precision medicine and imposes high demands on the accuracy of micro-biome analysis techniques. We propose an alignment- and reference- free subsequence based 16S rRNA data analysis, as a new paradigm for micro-biome phenotype and biomarker detection. Our method, called DiTaxa, substitutes standard operational taxonomic unit (OTU)-clustering by segmenting 16S rRNA reads into the most frequent variable-length subsequences. We compared the performance of DiTaxa to the state-of-the-art methods in phenotype and biomarker detection, using human-associated 16S rRNA samples for periodontal disease, rheumatoid arthritis and inflammatory bowel diseases, as well as a synthetic benchmark dataset. DiTaxa performed competitively to the k-mer based state-of-the-art approach in phenotype prediction while outperforming the OTU-based state-of-the-art approach in finding biomarkers in both resolution and coverage evaluated over known links from literature and synthetic benchmark datasets. AVAILABILITY AND IMPLEMENTATION: DiTaxa is available under the Apache 2 license at http://llp.berkeley.edu/ditaxa. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Genomic variation and strain-specific functional adaptation in the human gut microbiome during early life.

The human gut microbiome matures towards the adult composition during the first years of life and is implicated in early immune development. Here, we investigate the effects of microbial genomic diversity on gut microbiome development using integrated early childhood data sets collected in the DIABIMMUNE study in Finland, Estonia and Russian Karelia. We show that gut microbial diversity is associated with household location and linear growth of children. Single nucleotide polymorphism- and metagenomic assembly-based strain tracking revealed large and highly dynamic microbial pangenomes, especially in the genus Bacteroides, in which we identified evidence of variability deriving from Bacteroides-targeting bacteriophages. Our analyses revealed functional consequences of strain diversity; only 10% of Finnish infants harboured Bifidobacterium longum subsp. infantis, a subspecies specialized in human milk metabolism, whereas Russian infants commonly maintained a probiotic Bifidobacterium bifidum strain in infancy. Groups of bacteria contributing to diverse, characterized metabolic pathways converged to highly subject-specific configurations over the first two years of life. This longitudinal study extends the current view of early gut microbial community assembly based on strain-level genomic variation.

Reproducible Colonization of Germ-Free Mice With the Oligo-Mouse-Microbiota in Different Animal Facilities.

The Oligo-Mouse-Microbiota (OMM12) is a recently developed synthetic bacterial community for functional microbiome research in mouse models (Brugiroux et al., 2016). To date, the OMM12 model has been established in several germ-free mouse facilities world-wide and is employed to address a growing variety of research questions related to infection biology, mucosal immunology, microbial ecology and host-microbiome metabolic cross-talk. The OMM12 consists of 12 sequenced and publically available strains isolated from mice, representing five bacterial phyla that are naturally abundant in the murine gastrointestinal tract (Lagkouvardos et al., 2016). Under germ-free conditions, the OMM12 colonizes mice stably over multiple generations. Here, we investigated whether stably colonized OMM12 mouse lines could be reproducibly established in different animal facilities. Germ-free C57Bl/6J mice were inoculated with a frozen mixture of the OMM12 strains. Within 2 weeks after application, the OMM12 community reached the same stable composition in all facilities, as determined by fecal microbiome analysis. We show that a second application of the OMM12 strains after 72 h leads to a more stable community composition than a single application. The availability of such protocols for reliable de novo generation of gnotobiotic rodents will certainly contribute to increasing experimental reproducibility in biomedical research.

Transcriptome-wide analysis uncovers the targets of the RNA-binding protein MSI2 and effects of MSI2's RNA-binding activity on IL-6 signaling.

The RNA-binding protein Musashi 2 (MSI2) has emerged as an important regulator in cancer initiation, progression, and drug resistance. Translocations and deregulation of the MSI2 gene are diagnostic of certain cancers, including chronic myeloid leukemia (CML) with translocation t(7;17), acute myeloid leukemia (AML) with translocation t(10;17), and some cases of B-precursor acute lymphoblastic leukemia (pB-ALL). To better understand the function of MSI2 in leukemia, the mRNA targets that are bound and regulated by MSI2 and their MSI2-binding motifs need to be identified. To this end, using photoactivatable ribonucleoside cross-linking and immunoprecipitation (PAR-CLIP) and the multiple EM for motif elicitation (MEME) analysis tool, here we identified MSI2's mRNA targets and the consensus RNA-recognition element (RRE) motif recognized by MSI2 (UUAG). Of note, MSI2 knockdown altered the expression of several genes with roles in eukaryotic initiation factor 2 (eIF2), hepatocyte growth factor (HGF), and epidermal growth factor (EGF) signaling pathways. We also show that MSI2 regulates classic interleukin-6 (IL-6) signaling by promoting the degradation of the mRNA of IL-6 signal transducer (IL6ST or GP130), which, in turn, affected the phosphorylation statuses of signal transducer and activator of transcription 3 (STAT3) and the mitogen-activated protein kinase ERK. In summary, we have identified multiple MSI2-regulated mRNAs and provided evidence that MSI2 controls IL6ST activity that control oncogenic signaling networks. Our findings may help inform strategies for unraveling the role of MSI2 in leukemia to pave the way for the development of targeted therapies.

EDEN: evolutionary dynamics within environments.

SUMMARY: Metagenomics revolutionized the field of microbial ecology, giving access to Gb-sized datasets of microbial communities under natural conditions. This enables fine-grained analyses of the functions of community members, studies of their association with phenotypes and environments, as well as of their microevolution and adaptation to changing environmental conditions. However, phylogenetic methods for studying adaptation and evolutionary dynamics are not able to cope with big data. EDEN is the first software for the rapid detection of protein families and regions under positive selection, as well as their associated biological processes, from meta- and pangenome data. It provides an interactive result visualization for detailed comparative analyses. AVAILABILITY AND IMPLEMENTATION: EDEN is available as a Docker installation under the GPL 3.0 license, allowing its use on common operating systems, at http://www.github.com/hzi-bifo/eden. CONTACT: alice.mchardy@helmholtz-hzi.de. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Characterisation of a stable laboratory co-culture of acidophilic nanoorganisms.

This study describes the laboratory cultivation of ARMAN (Archaeal Richmond Mine Acidophilic Nanoorganisms). After 2.5 years of successive transfers in an anoxic medium containing ferric sulfate as an electron acceptor, a consortium was attained that is comprised of two members of the order Thermoplasmatales, a member of a proposed ARMAN group, as well as a fungus. The 16S rRNA identity of one archaeon is only 91.6% compared to the most closely related isolate Thermogymnomonas acidicola. Hence, this organism is the first member of a new genus. The enrichment culture is dominated by this microorganism and the ARMAN. The third archaeon in the community seems to be present in minor quantities and has a 100% 16S rRNA identity to the recently isolated Cuniculiplasma divulgatum. The enriched ARMAN species is most probably incapable of sugar metabolism because the key genes for sugar catabolism and anabolism could not be identified in the metagenome. Metatranscriptomic analysis suggests that the TCA cycle funneled with amino acids is the main metabolic pathway used by the archaea of the community. Microscopic analysis revealed that growth of the ARMAN is supported by the formation of cell aggregates. These might enable feeding of the ARMAN by or on other community members.

Genome-guided design of a defined mouse microbiota that confers colonization resistance against Salmonella enterica serovar Typhimurium.

Protection against enteric infections, also termed colonization resistance, results from mutualistic interactions of the host and its indigenous microbes. The gut microbiota of humans and mice is highly diverse and it is therefore challenging to assign specific properties to its individual members. Here, we have used a collection of murine bacterial strains and a modular design approach to create a minimal bacterial community that, once established in germ-free mice, provided colonization resistance against the human enteric pathogen Salmonella enterica serovar Typhimurium (S. Tm). Initially, a community of 12 strains, termed Oligo-Mouse-Microbiota (Oligo-MM12), representing members of the major bacterial phyla in the murine gut, was selected. This community was stable over consecutive mouse generations and provided colonization resistance against S. Tm infection, albeit not to the degree of a conventional complex microbiota. Comparative (meta)genome analyses identified functions represented in a conventional microbiome but absent from the Oligo-MM12. By genome-informed design, we created an improved version of the Oligo-MM community harbouring three facultative anaerobic bacteria from the mouse intestinal bacterial collection (miBC) that provided conventional-like colonization resistance. In conclusion, we have established a highly versatile experimental system that showed efficacy in an enteric infection model. Thus, in combination with exhaustive bacterial strain collections and systems-based approaches, genome-guided design can be used to generate insights into microbe-microbe and microbe-host interactions for the investigation of ecological and disease-relevant mechanisms in the intestine.