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

Prof. Dr.

Prof. , Dr.

BACAI: A novel multi-modal approach to bacterial image analysis and classification

Computational microbiology

The rapid and accurate identification of bacteria remains a critical challenge in microbiology, with far-reaching implications for healthcare, environmental monitoring, and industrial applications. We present BACAI (Bacterial Analysis and Classification through Artificial Intelligence), an innovative multi-modal platform designed to revolutionize bacterial image analysis and classification.BACAI employs a multi-modal approach, integrating various data types to enhance bacterial analysis. As a foundational step, we are comparing state-of-the-art segmentation models, including convolutional neural networks and U-Net architectures, to establish a performance baseline for fluorescence microscopy image analysis. This comparison will inform the development of our multi-modal framework, which aims to synthesize image data with other relevant bacterial characteristics.The platform's multi-modal capabilities extend beyond simple classification, aiming to extract meaningful biological insights from image data. By incorporating additional imaging modalities and metadata, BACAI has the potential to elucidate microbial cell biology characteristics, including morphological variations, growth patterns, and potentially, antibiotic resistance profiles.Our initial database focuses on three clinically relevant bacterial species: Escherichia coli, Staphylococcus aureus, and Pseudomonas aeruginosa. These species serve as a starting point for our model development, with plans to expand to a broader range of bacterial species in subsequent phases. This approach allows us to refine our methodologies on well-characterized organisms before tackling more diverse microbial communities.Our ongoing research focuses on optimizing the neural network architecture, expanding the training dataset, and developing robust data augmentation techniques to enhance the model's generalization capabilities. Future directions include exploring the platform's potential for analyzing microbial interactions within mixed populations and integrating genomic data to provide a more comprehensive understanding of bacterial physiology and behavior.BACAI represents a significant step towards automated, high-throughput bacterial analysis, with potential applications ranging from clinical diagnostics to environmental monitoring. This presentation will outline the theoretical framework of BACAI discuss the challenges, and present our roadmap for advancing the field of AI-assisted microbiology.

Poster

BakRep DB release 2.0

Bacteria serve as intriguing subjects of study across various disciplines for numerous reasons, and whole-genome sequencing has emerged as the primary method for enhancing our understanding of microbiology. The advent of cost-effective sequencing technologies has significantly accelerated bacterial whole-genome sequencing, presenting new challenges related to data accessibility, computational requirements, the diversity of analysis workflows, and ultimately, its scientific applicability. To address these challenges, Blackwell et al. released a consistently processed dataset of 661,405 bacterial genome assemblies sourced from the European Nucleotide Archive (ENA) as of November 2018. Building on this foundation, we utilized this dataset to compute assembly metrics, perform robust taxonomic classifications and MLST subtypings, and annotate all genomes. Previously, we introduced BakRep, a searchable large-scale web repository featuring this initial set of genomes, with plans to continuously expand the database by adding more genomes over time. In March 2024, as a follow-up to Blackwell"s initial dataset, the AllTheBacteria project released an update with a total of 1,932,812 assemblies. We processed all these additional genome assemblies using our previous workflow and made them accessible through our web repository. BakRep is an innovative web repository designed to improve the findability and accessibility of sequenced bacterial genomes housed in public data repositories. By offering a uniformly processed and thoroughly annotated dataset, BakRep tackles key challenges in the field of microbiology, such as data integration, standardization, and usability. We believe that this large-scale yet accessible genome repository will empower microbiology researchers across various disciplines to leverage this extensive collection of genomes as a foundation for their targeted analyses. With its user-friendly interface and comprehensive dataset, BakRep serves as an invaluable resource for the scientific community, facilitating a wide array of comparative and clinical studies and fostering new breakthroughs in microbial genomics.

Prof.

NFDI4Microbiota supporting microbiome research providing data access, services, training and workflows

Introduction: NFDI4Microbiota aims to support the microbiome research community by providing access to data, analysis services, data/metadata standards, and training. It belongs to the National Research data Infrastructure (NFDI), which aims to develop a comprehensive research data management. Different consortia ensure a broad coverage from cultural sciences, engineering to life sciences and natural science. NFDI4Microbiota intends to facilitate the digital transformation in the microbiological community (bacteriology, virology, mycology, and parasitology).Goals: NFDI4Microbiota aims to support the German microbiome research network through training and community building activities, and by creating a cloud-based system that will make the storage, integration, and analysis of microbial data and (microbial) omics data, consistent, reproducible, and accessible. Thereby, NFDI4Microbiota will promote the FAIR (Findable, Accessible, Interoperable and Re-usable) principles and Open Science. Results: To enable FAIR data management, the NFDI4Microbiota consortium develops and provides computational infrastructure and analytical workflows to store, access, process, and interpret various microbiome-related data types. NFDI4Microbiota works on developing and implementing software and standardized workflows for users to analyze their own data. Further, NFDI4Microbiota offers trainings, spanning from metagenomics, over courses about programming in R, to research data management and ELN (electronic lab notebooks). To interact with young scientists, the consortium launched an ambassador program, thereby helping to identify the needs of their local community. All relevant information and specific services are available via the web portal.Summary: NFDI4Microbiota has established community services providing access to data, analysis services, data/metadata standards, and training thereby promoting FAIR principles and Open Science in the microbiology community.

Thermodynamic limits of microbial growth

Microbial growth is often described as an energy converter, in which the catabolic breakdown of nutrients yields free energy which is partly used to drive the anabolic processes that assemble nutrients into new biomass. The combined effect of anabolism and catabolism can be described by a so-called "macrochemical" equation, which summarizes the chemical conversion of nutrients into catabolic waste products, biomass, and possibly other side products. Because such a description of microbial growth completely ignores the details of intracelleular metabolism, it is often termed a "black box" model of microbial metabolism. A diametrically different approach to describe metabolism is by genome-scale metabolic network models. Such models are derived from annotated genome sequences and ideally comprise all known biochemical reactions that can occur inside an organism. We here investigate how these two concepts can be combined to study the thermodynamics of microbial metabolism. We illustrate how experimentally determined macrochemical equations can be systematically separated into a catabolic and an anabolic part. We then characterize these separate chemical equations by their thermodynamic properties and derive the generalized thermodynamic forces and flows that correspond to the affinities of catabolism and anabolism, and the corresponding catabolic and anabolic rates. With these data we challenge the common view that microbial energy conversion can be described by a linear energy converter model. Moreover, we employ genome-scale models to study how catabolism and anabolism are coupled through the production and consumption of ATP, and determine thermodynamic efficiencies. Finally, we observe a linear relation between the growth rate and the catabolic power, which appears to be species independent and may hint at a general principle in microbial growth.

Pan-BGC Atlas: An in-depth visual analysis of biosynthetic gene cluster families

Due to environmental pressures such as microbial competition and adaptation to diverse ecological niches, biosynthetic gene clusters (BGCs) in microorganisms have evolved a remarkable natural diversity. This diversity enables microbes to produce a wide array of structurally and functionally similar yet distinct secondary metabolites, many of which play essential roles in survival and interaction within their environments.Using bioinformatics tools, we can group BGCs into gene cluster families (GCFs) based on genetic similarity. GCFs can vary widely in size, ranging from single clusters per family to hundreds. Compounds produced by BGCs within a family typically share a core chemical structure with only slight variations. This raises important questions: What are the underlying mechanisms and building blocks that lead to this diversity? Can we harness this potential, and gather more insights into selection mechanisms?To better understand and systematically explore the natural diversity in GCFs, we developed a comprehensive visual GCF database, the Pan-BGC Atlas, by adapting the pangenome concept to BGCs. Our approach involved grouping bacterial BGCs from the antiSMASH database into similarity-based families, followed by a series of bioinformatics analyses. In the Pan-BGC Atlas, each GCF is visually represented and can be easily adjusted, allowing for rapid and intuitive analysis.The Atlas includes both an "open" BGC, encompassing all possible genes within a GCF, and a "closed" BGC, representing the minimal set of genes needed to produce the core chemical structure shared across the family. Comparing the open and closed BGCs, along with other information, reveals enzymatic diversity and could offer new insights for synthetic biology.Beyond exploration, users can compare their own BGCs against the database to identify the closest matching families. Additionally, a downloadable pipeline enables users to create similar visual analyses for their own gene cluster families, fostering reproducibility and broader application of these tools. This platform is a valuable resource for microbial genomics and natural product research, enhancing navigation and interpretation within GCFs.

Importance of data Provenance in genome sequencing studies. A guideline to understand the gaps in sequencing projects

There is a growing need to make data from life sciences findable, accessible, interoperable, and reusable(FAIR), which is incredibly challenging in metagenomics. This field deals with the study of microorganisms' functional potential in natural, host-associated, and constructed environments. It uses high-throughput sequencing data from total DNA isolated from microbial communities. There is often a disconnection between wet labs and computational analysis in this field, leading to provenance issues among scientists.This study aims to improve FAIR principle usage in metagenomics by creating guidelines to connect the wet and dry lab parts of projects in the context of genomic sequencing. To this end, we explored provenance in over 50 projects involving more than 1000 samples from multiple sources, i.e., bioreactors, agricultural and forest soils, freshwater, wastewater, and the gut microbiome. We organized the guidelines into five parts: sample preparation, sequencing logistics, data downloading, integrity and quality checking of sequences, and preprocessing the sequenced data until assembly.First, we analyzed provenance from sample preparation (sample storage and DNA extraction) for metagenomics. Our data indicated that it is necessary to pay attention to the DNA yield and quality (particularly if interested in long-read sequencing) during sample preparation for sequencing. We observed that both low and high yields in the samples can lead to failed library preparation. DNA quality check must be performed in every sample in BioAnalyzer-like machines. We also highlight that memory configurations and resources required for sequencing projects to facilitate reproducibility must be considered, particularly for sequencing data preprocessing, as it demands a high usage of resources. Reviewing and commenting must be implemented on automatic reports in existing data processing pipelines to improve the interoperability and reuse of metagenomics data. In Conclusion, small research groups and data stewards responsible for organizing data in local or large sequencing facilities may use our guidelines to bridge the gap between wet and dry lab researchers. Our policies may help to improve FAIR usage of metagenomics as we concentrate on data interoperability and reuse aspects in genome sequencing.

Dr.-Ing.

Extending the enzyme-constrained genome scale model of Corynebacterium glutamicum with multi-omics datasets

IntroductionAdvances in computational methods continue to shift the landscape of what is feasible when modeling complex biological systems in-silico, while modern omics analysis tools can deliver the broad in-vivo quantification of biomolecules required to calibrate these models. Genome-scale metabolic models (GEMs) represent the entirety of known biochemical reactions encoded in a genome as a network of constrained and balanced fluxes. Several GEM extensions have been developed, integrating details from systems biology (e.g. transcriptional regulation [1]) and biochemistry (e.g. turnover numbers [2]).GoalsThis work focuses on a novel enzyme-constrained GEM (ecGEM) for C. glutamicum integrating data both from systems biology and biochemistry.Materials & MethodsThe ecGEM formulation presented here combines (i) enzyme abundances derived from proteomics with (ii) catalytic turnover numbers estimated via data mining and machine learning methods based on the public databases BRENDA and SABIO-RK as well as the (iii) incorporation of transcriptomics derived from RNA-seq.ResultsVia data mining, 457 turnover numbers could be retrieved for the catalytic enzymes of C. glutamicum (30% of 1539). The remaining 70% were predicted via recently published deep learning models. Protein and RNA abundances were determined across a broad set of culture conditions. In general, more than 1000 proteins were quantified via LC-MS-MS with data-independent acquisition, while the total RNA profile was investigated with RNA-seq. Those datasets served for the calibration of the newly developed ecGEM. Computational methods were applied for the prediction of protein abundances from mRNA, enabling the direct use of transcriptome profiles as condition-specific constraints for the analysis of the ecGEM.SummaryThe resulting ecGEM can be used to investigate C. glutamicum on the transcriptomic, proteomic and metabolic level, providing new impetus to metabolic engineering as well as a proof of concept for the extension of ecGEMs beyond the proteome.References [1]: E. Motamedian et al., "TRFBA: an algorithm to integrate genome-scale metabolic and transcriptional regulatory networks with incorporation of expression data," Apr. 2017, doi: 10.1093/bioinformatics/btw772.[2]: T. B. Alter, L. M. Blank, and B. E. Ebert, "Proteome Regulation Patterns Determine Escherichia coli Wild-Type and Mutant Phenotypes," Mar. 2021, doi: 10.1128/msystems.00625-20.

Computational microbiology 2

Termin

Programm