7th joint conference "Joint Microbiology & Infection Conference 2024" of the DGHM and VAAM

Chair

Professor

LLM-based question answering for the infectious disease domain

Microbiology in the digital Era

IntroductionUsing inappropriate antibiotics can quickly lead to antimicrobial resistance and infections that are difficult to treat. However, the capacity for training expert medical personnel is very limited, especially in developing countries. To ameliorate this gap, we develop a personalized AI-supported tutor for antibiotic usage training, with a chatbot as one of its core components.MethodologyAt the heart of the chatbot is a set of large language models (LLMs). However, LLMs are known to &quot;hallucinate&quot;, i.e., generate factually incorrect information. To tackle this, we use a retrieval augmented generation (RAG) approach that restricts the LLM response to the evidence context retrieved from a curated corpus of credible data. In our case, we collected an intentionally diverse set of media from guidelines and (review) articles to podcasts and blog posts across several levels of evidence quality.Our retriever combines lexical retrieval models (e.g., BM25), re-ranking based on the question–document semantic similarity, and topical routing. The routing filters out retrieved documents whose modeled topic mismatches the question topic (e.g., &quot;skin infection&quot;).To tackle complex questions (e.g., &quot;Which empiric therapy should be used for erysipelas in a patient with kidney failure and diabetes mellitus type 2?&quot;), we implement question decomposition that creates a set of shorter queries (e.g., &quot;empiric therapy for erysipelas&quot;, &quot;therapy with kidney failure&quot;, etc.).Furthermore, we add the answer fact verification and source evidence citation modules. The fact verification module ensures that the generated answer&quot;s statements are actually supported by the retrieved evidence, and evidence citation allows the user to verify the statements by inspecting the evidence sources.SummaryWe briefly describe our ongoing work on developing an AI-supported chatbot that assists in the rational selection of antibiotic therapy. The chatbot architecture combines RAG with fact verification and evidence citation modules. The planned evaluation will reveal existing drawbacks and allow us to address them.

Oral Presentation

Prof. Dr.

StrainRegistry – A central registry for microbial strains

Increasingly, large microbiome studies generate considerable numbers of isolates. Deposition of these isolates to culture collections is not only in the interest of open science but is also required to formally name the many new taxa usually found among them. This presents a challenge to both researchers and biological resource centers. While the former wish to receive culture collection numbers as soon as possible to use as persistent identifiers in publications, the latter do not have the capacity to process these sizable collections simultaneously.Here we present StrainRegistry, a novel strain registration and deposition management service connected to the StrainInfo database (https://straininfo.dsmz.de). It allows microbiologists to register strains with strain identity information to be published as strain entries in the StrainInfo database. These can be linked with culture collection numbers, sequences and publications as they become available. Each strain entry receives a persistent StrainInfo identifier and its respective DOI. While this cannot replace deposition in a culture collection and the rigorous quality checks needed, it will allow for consistent referencing of the strain from the very outset.StrainRegistry also aims to simplify the deposition process by giving researchers the option to submit further meta- and deposition data, which are not published, but can directly be shared with a participating culture collection through the StrainRegistry deposition manager. This user interface visualizes the deposition status of each strain and facilitates communication between the depositor and the responsible culture collection curator, who can view strain information, demand further data, request shipping of biological material and change strains statuses.This procedure is being developed in collaboration with the curators of the DSMZ to simplify bulk deposition at the DSMZ and with the ambition to engage other culture collections and standardize microbial strain deposition worldwide.

Jun.-Prof. Dr.-Ing.

Detecting interpretable patterns in estuarine-coastal microbial communities using methods from computational linguistics and Machine Learning

Aquatic bacterial communities are integral components of earths ecosystems. Investigating temporal and spatial community patterns provide insights into the dynamics of the entire ecosystem. To uncover such patterns, next-generation sequencing methods provide the necessary data. But these data sets are usually compositional, extremely high-dimensional and exhibit non-linear relationships, which makes them difficult to understand and to analyse using statistical methods.To circumvent these problems and to identify understandable compositional patterns in their temporal and spatial dynamics, we applied Topic Modeling (TM), a method originally from computational linguistics, to the microbial communities of the Warnow estuary and the Baltic Sea coast. This environment is characterized by strongly fluctuating environmental conditions and a steep salinity gradient and was sampled in high spatio-temporal resolution between April 2022 and April 2023.More specifically, we compared several TM algorithms as well as various data preprocessing methods in their ability to capture ecological relationships using a Random Forest-based machine learning approach. This way, we could verify how much information was retained through TM while minimizing the dimensionality of the feature space as much as possible.Using this approach, we identified subcommunities (corresponding to topics) specific to freshwater and brackish environments and specific seasonal phases. These subcommunities can be linked across primer sets and, therefore, the domains of life. In the estuary, we are able to link the retention time of specific subcommunities to weather and hydrodynamical dynamics.Based on these results, we conclude that TM achieves a dimensional reduction of microbiome datasets that significantly increases their comprehensibility. Taking advantage of structural similarities between linguistic and microbiome datasets, TM proves to be a powerful tool especially for the analysis of microbial communities in dynamic habitats.

Decoding Antibiotic Resistance: OprD Allels in Pseudomonas aeruginosa and their Impact on Carbapenem Resistance

Pseudomonas aeruginosa, commonly isolated from hospitalized patients, poses a significant therapeutic challenge due to its intrinsic and acquired antibiotic resistance. Carbapenems, crucial in treatment, often fail to eradicate the pathogen due to resistance associated with mutations in the outer membrane porin D (OprD). This study seeks to elucidate the impact of distinct OprD alleles (amino acid sequences) on carbapenem resistance, aiming to establish OprD as a reliable genetic predictor.A dataset comprising 4,875 clinical isolates collected unbiased from a German and a Danish hospital, underwent susceptibility testing against imipenem and meropenem and whole genome sequencing. The analyzed isolates contained 192 distinct OprD alleles, with the most common variant occurring in 20.7 % of the isolates. Of the 4,875 isolates 46.7 % decoded an OprD sequence of conserved-length (441-446 amino acids). There were in total 41 different conserved-length Oprd variants. The fact that a clinical P. aeruginosa isolates does not contain a conserved-length allele showed unexpectedly high accuracy in predicting the antimicrobial resistance for both imipenem (sensitivity: 0.731; specificity: 0.962) and meropenem (0.764; 0.934). Further analysis focused on common OprD variants that were found in at least 10 clinical isolates (30 OprD variants) and revealed significant differences in their individual impact on carbapenem resistance. Interestingly, there are also variants that are exclusively predictive for antimicrobial resistance of either imipenem or meropenem.The results again underline that OprD is the most common driver of carbapenem resistance in P. aeruginosa in terms of non-acquired resistance and it is the first time that OprD alleles have been analyzed in &gt;1,000 isolates. We found that inactivation of OprD is usually caused by an early stop codon or a frameshift mutation and rarely by single amino acid exchanges.

Modeling microbial communities using biochemical resource allocation analysis

Introduction: Oxygenic photosynthesis, originating in cyanobacteria, is a critical biological process that drives primary production in most ecosystems. While we understand cyanobacterial growth dynamics in axenic laboratory cultures, most microorganisms evolved in interconnected ecosystems, and understanding their physiology requires taking their evolution within these dynamic ecosystems into account. Goals: Our objective is to understand the emergence of interactions between photo- and heterotrophic microorganisms using computational models grounded in biochemical resource allocation analysis. We aim to delineate the prerequisites and energetic trade-offs governing cooperation, division of labor, and nutrient cycles in microbial communities. Materials &amp; Methods: Building on previous advancements, we make use of high-quality quantitative computational models of microbial growth and resource allocation. Specifically, we simulate co-cultures of photo- and heterotrophic organisms, where each microbial partner maximizes its growth rate, allowing us to explore whether interactions and dependencies emerge. Results: Using a novel computational framework, we simulate interactions between photo- and heterotrophic microorganisms. By examining the costs and benefits of these interactions, we outline a plausible evolutionary pathway for the emergence of metabolic dependencies between these microorganisms in marine environments. We show that co-cultures can result in long-term stable cultures with increased productivity compared to axenic growth. Summary: The perspective of cellular resource allocation offers a unique opportunity to understand the constraints and energetic trade-offs that govern the emergence of dependencies between photo- and heterotrophic microorganisms.

Development of an integrated genomic surveillance (IGS) of public health relevant pathogens in Germany

Integrated genomic surveillance (IGS) is an effective public health strategy for the surveillance of infectious pathogens, which is becoming increasingly important worldwide. IGS combines the results of DNA sequencing and bioinformatics genome analyses of pathogen isolates from patient samples with epidemiological information from the reporting system and further data.The main objective of IGS is to rapidly detect spread of infectious agents and to identify links, even between agents detected at different times and locations. The combination of genomic pathogen sequence data with selected epidemiological information on infected individuals allows outbreaks and transmission chains to be identified quickly and targeted interventions to contain the outbreak to be initiated at an early stage. In addition, the IGS offers the possibility of continuous detection and surveillance of pathogen variants with changing characteristics (e.g. transmissibility, virulence, AMR). The IGS also makes an important contribution to the decision-making process of public health authorities.Two projects funded by the Federal Ministry of Health are currently underway with the aim to establish an IGS at the RKI in cooperation with the National Reference Centres (NRZ) and the Consiliary Laboratories (KL) as well as other stakeholders such as the health authorities of the federal states and the Network University Medicine (NUM). The necessary processes and an efficient infrastructure will be developed and established for the long-term implementation of the IGS according to §13 of the Infection Protection Act.The necessary framework conditions for the systematic sequencing of infectious pathogens in the corresponding NRZ/KL have been created. The expansion of the German Electronic Reporting and Information system (DEMIS) for the transmission of genomic data is underway, as is the establishment of a high-performance IT infrastructure for the genome analysis, linking with epidemiological data, visualisation and provision of the information to stakeholders.The so far established IGS nucleus should be stringently developed into a comprehensive sustainable Public Health Instrument in Germany.

Oral Session
Microbiology in the Digital Era

Termin

Chair

Thema

Programm

Oral SessionMicrobiology in the Digital Era

Termin

Chair

Thema

Programm

Oral Session
Microbiology in the Digital Era