Annual Conference 2025 of the Association for General and Applied Microbiology - Jahrestagung 2025 der Vereinigung für Allgemeine und Angewandte Mikrobiologie

Chair

Prof.

Towards predicting significant changes in wastewater microbial communities: A machine learning approach using time series data

Computational microbiology

During the Covid-19 pandemic, wastewater-based surveillance has emerged as a pivotal technology for monitoring health-related targets in the public sector [1]. We seek to support the monitoring of the various bacterial genera present in wastewater by developing a machine learning-based early warning system capable of detecting possible changes posing a threat to human health. To this end, the prediction performance of different machine learning architectures for bacterial time series data with and without additional metadata was tested. While the prediction of bacterial abundances returned satisfactory results, it became evident that the added metadata needed to be carefully selected depending on the sampling environment. Thus far, Long Short-Term Memory (LSTM) [2] models have demonstrated the most promising prediction performance. A prediction interval was constructed to encompass typical fluctuations in the data. Feature importance analysis was conducted using Shapley Additive Explanations (SHAP) [3]. In the subsequent stages, we intend to integrate outlier detection to develop an early warning system and investigate the potential of transfer or ensemble learning to enhance the model's performance and its applicability to other datasets.[1] D"Aoust, P. et al. SARS-CoV-2 viral titer measurements in Ontario, Canada wastewaters throughout the COVID-19 pandemic. 2024. Scientific Data.[2] Baranwal, M. et al. Recurrent neural networks enable design of multifunctional synthetic human gut microbiome dynamics. 2022. eLife.[3] Lundberg, S.M., Lee S.-I. et al. A Unified Approach to Interpreting Model Predictions. Part of NIPS 2017.

Short lecture

Assessment and prediction of spatio-temporal dynamics of Vibrio vulnificus in the coastal Baltic Sea

With rising infection rates in recent years Vibrio vulnificus pose an increasing threat to public safety in the coastal brackish Baltic Sea. V. vulnificus, associated with high mortality, can thrive in the Baltic Sea during summer months due to optimal conditions and can cause severe infections to humans through open-wound infections. Pinpointing a timeframe of increased risk of infection is the next step in curbing the increasing numbers of infections and also has wide-ranging ecological and economic effects. Routine monitoring of this bacterium in the Baltic Sea is critical to provide a warning system for the public when the risk of infection is potentially high. As part of an extensive twice weekly sampling campaign that included 14 locations in the coastal Baltic Sea across a one-year period between 2022-2023, we investigated the abundance of V. vulnificus using a multi-method approach, including droplet digital PCR (ddPCR) results targeting the vvhA gene region, agar cultivation, and species level classification of 16S rRNA gene sequencing. Physico-, biological- and hydrochemical parameters were measured concurrently and variables explaining V. vulnificus occurrence were identified by machine learning. In addition, numerous machine learning techniques were applied in order to predict V. vulnificus gene concentrations using ddPCR, with the objective of establishing models that accurately predict when V. vulnificus is most abundant in the coastal Baltic Sea. Time-series analysis was performed using variables from previous timepoints as predictors, with the goal of pinpointing the most important markers for creating an early warning system and highlighting the importance of improved coastal monitoring. First results will be presented at the conference.

Prof. Dr.

Toxin-Antitoxin Systems – How does the double-edged sword look like?

Antibiotic resistant bacteria are an immediate threat to public health, animal husbandry and agriculture. Toxin-antitoxin (TA) systems in multidrug-resistant bacteria are promising approaches to combat them, but further research is needed to fully understand their distribution, functioning, regulation and evolution. TA systems are genetic modules consisting of a toxin and its antitoxin, which are proteins or rnas, and are involved in various cellular processes, including cell death and stress response. Eight different types, with type I and II being the most investigated, are known.This project aims to computationally identify and characterize all TAS in all bacteria, archaea and phages from NCBI to gain more insights by using sequence similarity searches and hidden markov models and in the case of ncRNA component, we use RNAalifold1, Infernal2 and R-scape3. In contrast to a simple search, which is limited to direct sequence and structural similarity, our iterative approach allows us to identify more evolutionary divergence TA systems and enhances both sensitivity and accuracy. Furthermore, we intend to employ a single-plasmid-based approach to experimental verify novel TA systems of interest.Through sequence alignment and secondary structure prediction, we identify conserved motifs and structural elements within these RNA components. Additionally, we explore potential RNA-RNA and protein-RNA interactions. Our findings provide insights into the structural and functional characteristics of these systems on a global level, paving the way for further experimental validation. We'll also expand the RFAM database to aid future research.A deeper understanding of TA systems is an essential part of the search for alternatives to antibiotic treatment and future applications in medicine, animal husbandry, and agriculture. A comprehensive search shall reveal their whereabouts and appearance.1 Bernhart, Stephan H., et al. https://doi.org/10.1186/1471-2105-9-4742 Nawrocki, Eric P., and Sean R. Eddy. https://doi.org/10.1093/bioinformatics/btt5093 Rivas, Elena, Jody Clements, and Sean R. Eddy. https://doi.org/10.1038/nmeth.4066

PD Dr.

Resolving challenges in Nanopore-Based sequencing and assembly of bacteriophage genomes

Genome sequencing and assembly present challenges, with short-read methods often leading to fragmented assemblies, while long-read methods, though capable of providing contiguous reads, can lack accuracy. Recent advances in Nanopore's "duplex" technology improved per-base quality for long-reads, making it promising for genome assembly. To evaluate this for smaller genomes, we resequenced 182 bacteriophages, ranging from 6 to 79 Kb, and attempted assembly using state-of-the-art tools and best practices available. Only up to 15.4% (29 out of 182) of the initial assemblies matched the expected genome size (±5% bp). Our analysis revealed three key obstacles: (1) inefficient demultiplexing resulting in fewer reads binned, (2) barcode bleeding introducing contaminating reads from different samples, and (3) genomic concatemers generating reads exceeding the actual genome size. To address these, we developed two tools, "Barbell" and "NanoPrepPhage"; the former improves demultiplexing, and the latter removes contaminated reads and reduces the impact of concatemers. Unlike existing tools, Barbell does not rely on the edit distance but instead uses the expected Nanopore current to demultiplex reads. NanoPrepPhage filters contaminant reads using K-mer profiling, and trims reads to a shorter length (either 100, 250, or 500 bps), reducing the impact of concatemers. Combining both tools, all genomes were assembled with their expected sizes, highlighting the robustness and effectiveness of our protocol. I will explain the implementation of the tools, present our findings in bacteriophage sequencing, and detail the limitations we encountered. Combining Barbell with NanoPrepPhage together provides a new best practice that fully exploits the benefits of Nanopore "duplex" reads for reliable genome assembly.

PhyloNaP: a tool for phylogeny-based function prediction of enzymes involved in biosynthesis of natural products

Introduction: The rise of antibiotic resistance underscores the urgent need for novel antibiotics derived from natural products (NPs). Genome mining has revealed diverse biosynthetic gene clusters (BGCs) that encode these compounds. Accurate structural knowledge of NPs is essential for extraction and other experimental procedures; however, structure prediction is challenging due to limited functional annotations for tailoring enzymes in NP biosynthesis.Goals: This project aims to develop PhyloNaP, a database and tool for the functional annotation of enzymes involved in NP biosynthesis, with the focus on tailoring enzymes using phylogenetic analysis. By consolidating fragmented phylogenetic and functional data, PhyloNaP enables more accurate NP structure predictions, ultimately aiding in the discovery of new antibiotics and other therapeutics.Materials & Methods: PhyloNaP integrates functional annotations from literature and public repositories such as MiBiG, SwissProt, and AntiSMASHdb. The database stores phylogenetic trees, sequence alignments, and functional annotations gathered from publications and generated specifically for this project. Currently, PhyloNaP includes enzyme families like methyltransferases, P450s, and halogenases, with each tree annotated with experimentally validated data on substrates and products.Results: Using PhyloNaP, researchers can explore annotated enzymes, classify new enzymes, and position them within existing phylogenetic trees. This mapping allows users to examine the functional characteristics of closely related enzymes. PhyloNaP also demonstrate substrate features and potential modification sites, enhancing the reliability of NP structure predictions. Additionally, PhyloNaP incorporates information on the taxonomy of the enzyme"s source organism and the main characteristics of its BGC, with links to corresponding NPs in the MiBIG database.Summary: PhyloNaP fills a critical gap in NP research by providing a centralized repository of annotated phylogenetic data. Researchers are encouraged to contribute new trees, thereby enriching this shared resource and increasing its utility across the scientific community. By integrating both phylogenetic and functional data, PhyloNaP has the potential to accelerate the discovery of valuable natural products.Acknowledgment: This project has received funding from the European Union"s Horizon Europe program under the Marie Skłodowska-Curie grant agreement No 101072485.

The de.NBI resource center for microbial bioinformatics

The German Network for Bioinformatics Infrastructure – de.NBI is a national, academic, and non-profit research infrastructure initiated in 2015 by the Federal Ministry of Education and Research. Within de.NBI, the Bielefeld-Gießen Resource Center for Microbial Bioinformatics (BiGi) supports scientists with bioinformatics software tools, pipelines and resources for microbial research with a focus on the analysis of microbial genomes, metagenomes, and related datasets.In the last decade, various software tools, pipelines and platforms for annotation, characterization and comparative analysis of genomic as well as metagenomic sequencing data have been maintained and further developed based on feedback from our growing user community. For the comprehensive annotation of microbial genomes, the fully-automated annotation software tool Bakta was developed. Bakta is available as command line tool as well as via a user-friendly web version. Bakta quickly emerged to be one of the most-actively used annotation tools worldwide. For the comparative analysis of bacterial genomes, we maintain the de.NBI flagship tool EDGAR which provides public as well as custom, password protected databases with more than 100,000 included genomes. Furthermore, a broad range of more specialized software tools were developed and maintained, e.g. Platon for the detection of plasmid-borne contigs within draft genomes, ReferenceSeeker for the taxonomic classification of microbial genomes, and MGX2 for the processing and analysis of metagenomic data. A further important project is the large-scale bacterial genome repository BakRep, which provides access to a comprehensive assembly and annotation data, enriched with taxonomic information, for a total of ~2.4 million genomes.Besides the development and maintenance of analysis software and data resources, BiGi also hosts a branch of the federated de.NBI cloud, which provides large-scale compute resources to the scientific community on demand. Additionally, BiGi offers training programs, workshops and user support to help the scientific community to use the various bioinformatics solutions effectively in their research.In summary, the microbial researcher community can strongly benefit from the triplet of services provided by BiGi:A broad set of well-maintained and constantly improved software solutionsAn accessible powerful and growing cloud-computing infrastructureA regular offer of accompanying workshops and training courses

Computational microbiology

Termin

Chair

Thema

Programm