Predicting the three-dimensional conformations of drug-like molecules is central to computational drug discovery. Molecular docking, pharmacophore modeling, shape-based virtual screening, and structure–activity analysis all depend on conformer ensembles, whose quality directly shapes downstream predictions. Current distance geometry methods such as ETKDGv3, implemented in the widely used RDKit cheminformatics library, rely on generic geometric priors and torsional preferences derived from crystallographic databases, limiting coverage and requiring ongoing manual curation. Here, we replace these empirical priors with geometric parameters derived from invariom model compounds computed at a semiempirical quantum-mechanical level of theory. These parameters are precomputed once and reused across target molecules, so no per-molecule quantum-mechanical calculation is required at conformer generation time. Coupling distance geometry with Monte Carlo sampling guided by invariom-based energy terms yields conformational ensembles that are systematically lower in energy than those generated by ETKDGv3, without the need for manually curated torsion rules. Because invariom parameters are computed rather than extracted from experimental data, this approach extends naturally to chemistries underrepresented in crystallographic repositories, providing a scalable route to improved conformer generation for drug discovery.

There has been an exponential increase in the development of technologies that have offered unprecedented opportunities to unravel the genetic underpinnings of rare and complex diseases. However, despite the progress, and availability of tools, there are still several challenges to overcome, such as the development of a unified framework that facilitates downstream analysis. To address this challenge, we present IDEAL-GENOM (Integrated Downstream Analytical Toolkit for Genomic Analysis), a Python-based wrapper designed to streamline the analytical workflow commonly implemented in genome-wide association studies (GWAS) settings. IDEAL-GENOM integrates widely used tools such as PLINK, GCTA and bcftools along with custom-developed functionalities, enabling reproducible results through parameter sharing. IDEAL-GENOM provides a simplified framework for quality control (QC) and GWAS and downstream analysis that will enable beginners and advanced users to leverage the in-built functionalities to perform GWAS analysis for complex diseases. IDEAL-GENOM is a customizable tool that enables users to efficiently analyze genotyping data with reproducible QC, post-imputation, and GWAS pipelines.

Antimicrobial resistance (AMR) surveillance is primarily based on clinical isolates, providing limited insight into population-level resistance dynamics. Metagenomic sequencing enables large-scale profiling of antibiotic resistance genes (ARGs), but integrating such data with heterogeneous socio-economic and ecological variables requires robust statistical and computational frameworks.
We analysed over 58,000 publicly available human gut metagenomes from NCBI and ENA, spanning more than 60 countries. ARG profiles were quantified by mapping reads to the ResFinder database, and microbiome composition was derived from marker-gene-based taxonomic annotation. These data were integrated with country-level indicators, including antibiotic consumption, socioeconomic indicators, and global connectivity.
We combined generalized additive models and Bayesian multilevel models to capture non-linear associations and region-specific variability while accounting for hierarchical structure. To explicitly model relationships between countries, we constructed dissimilarity matrices and apply dyadic regression to quantify how resistome differences correlate with geographic, socioeconomic, and microbiome distances and similarity.
ARG load and diversity are higher in low- and middle-income countries, with several predictors showing context-dependent or reversed associations across income groups. At the country level, resistome dissimilarity is strongly associated with microbiome composition and socio-economic similarity. Adjustment for microbiome structure reduces the strength of associations between socioeconomic distance and resistome variation, which suggests that microbiome composition partly accounts for the association between socio-economic factors and resistance patterns
These results show that global AMR patterns are structured by interacting socioeconomic and ecological factors, and demonstrate the value of combining hierarchical and dyadic modelling for analysing population-scale metagenomic data.

Active learning-guided docking enables efficient exploration of ultra-large chemical libraries otherwise inaccessible due to the cost of exhaustive screening. By iteratively selecting the most informative compounds, it prioritizes candidates with high predicted docking scores and model uncertainty, ensuring balanced exploration of chemical space. Only a small fraction of molecules is sampled, reducing computational demands while maintaining strong enrichment of promising candidates.
In this work, the method was applied to LasR, a key regulator of bacterial virulence. Targeting virulence factors such as LasR is particularly attractive because it may reduce the selective pressure associated with traditional antibiotics and limit the emergence of resistance. Moreover, the scarcity of known ligands for LasR makes structure-based approaches especially relevant.
Overall, active learning-guided docking enables scalable and flexible navigation of chemically diverse, billion-scale libraries, supporting the identification of novel drug candidates and providing a practical framework for integration with experimental validation workflows. Importantly, experimental validation of selected candidates confirmed the models predictive capability, highlighting its promise as an effective tool for accelerating drug discovery.
Importantly, the iterative nature of active learning allows continuous refinement of the model as new data is incorporated. This closed-loop workflow efficiently allocates computational resources to the most relevant regions of chemical space and bridges the gap between in silico screening and real-world drug discovery.

Acknowledgements: This work was supported by the National Science Centre (NCN), Poland, under the SONATA-19 grant no. 2023/51/D/NZ7/0259.

As Large Language Models (LLMs) transition from conversational assistants to autonomous research agents, their ability to interface with primary genomic databases becomes critical. However, the stochastic nature of LLMs is fundamentally at odds with the deterministic precision required for viral sequence retrieval. When navigating the multi-dimensional metadata of repositories like NCBI Virus, state-of-the-art agents (including GPT-5.5, Claude Sonnet 4, Edison and Biomni) frequently "hallucinate" query parameters or fail to navigate complex web-based filtering logic, leading to corrupted datasets and invalid downstream biological conclusions.

We present gget virus, a programmatic framework that formalizes the full filtering logic of the NCBI Virus portal into a deterministic system. To evaluate its impact, we developed VirBench, a curated benchmark of 120 complex virological queries. We demonstrate that unconstrained AI agents achieve a retrieval accuracy of only 16.9-91.3% with high inter-run variability. By equipping these agents with gget virus as a tool-use layer, accuracy increases to >90% across all systems, while ensuring exact-match semantics. Furthermore, the framework's "metadata-first" retrieval strategy reduces data transfer by >98% compared to bulk download methods. By bridging the gap between natural language intent and precise data acquisition, gget virus provides the necessary infrastructure for reproducible, AI-driven viral discovery and genomic surveillance. The manual and source code are available at https://scverse.org/gget/en/virus.html and https://github.com/scverse/gget/blob/main/gget/gget_virus.py, respectively.

RNA is increasingly recognized as a central regulator of diverse cellular processes, with non-coding mRNA regions encoding information that shapes subcellular localization, stability, and translation. This regulation cannot be explained by individual factors in isolation; it emerges from combinatorial interactions in which multiple RNA-binding proteins (RBPs) and RNA modifications co-occupy the same transcripts, assembling local interactomes that vary across transcript regions and cellular states. Mapping this landscape experimentally is constrained by the scale of the combinatorial space, the cost of specialized assays, and the context-specific nature of each measurement, leaving existing atlases confined to few contexts. Pretrained RNA Language Models (RNA-LMs) offer a complementary route: as shown by preliminary work in our group, even without modification labels their frozen embeddings recover modified nucleotides along with their canonical motifs, indicating they capture regulatory signal extractable from their representations.

We extend this paradigm to RBPs, treating the latent geometry of RNA-LM representations as a context-agnostic map of the regulatory landscape that encodes the implicit rules of post-transcriptional regulation. Probing this signal across model depth and sequence context quantifies how well the representation space captures known RBP binding preferences. Applied to a broad panel of RBPs profiled by CLIP-based assays across cellular contexts, the framework reveals a spectrum of recoverability. RNA-LM-derived predictions then feed transcriptome-wide non-negative matrix factorization (NMF), uncovering latent regulatory programs defined by co-occurring modifications and RBPs.

We present hecate, a modular lossless genomic compression framework. It is designed around uncommon but practical source-coding choices.
Unlike many single-method compressors, hecate treats compression as a conditional coding problem over coupled FASTA/FASTQ streams
(control, headers, nucleotides, case, quality, extras). It uses per-stream codecs under a shared indexed block container. Codecs include alphabet-
aware packing with an explicit side channel for out-of-alphabet residues, an auxiliary-index Burrows-Wheeler pipeline with custom arithmetic
coding, and a blockwise Markov mixture coder with explicit model-competition signaling. This architecture yields high throughput, exact
random-access slicing, and referential mode through streamwise binary differencing. In a comprehensive benchmark suite, hecate provides
the best compression vs. speed trade-offs against state-of-the-art established tools (MFCompress, NAF, bzip3, AGC), with notably stronger
behaviour on large genomes and high-similarity referential settings. For the same compression ratio, hecate is 2 to 10 times faster. When given
the same time budget as other algorithms, hecate achieves up to 5% to 10% better compression.

Microbial communities play critical roles across diverse biological systems, yet identifying the environmental factors that shape their composition remains challenging. Advances in sequencing technologies have increased data availability, intensifying the need for analytical approaches capable of handling high-dimensional, compositional microbiome data. While there is strong interest in uncovering multi-factor environmental drivers, conventional methods such as PERMANOVA, redundancy analysis (RDA), and Mantel tests rely on linear assumptions that often fail to capture non-linear, threshold-like patterns and complex interactions common in ecological systems.

Here, we propose an automated, machine learning–based framework to identify environmental drivers of microbial community structure in a user-friendly and generalizable manner. By relaxing strict model assumptions, this approach enables detection of complex, non-linear relationships between environmental variables and community composition.

Simulation studies using synthetic microbial communities demonstrate that the proposed machine learning framework outperforms traditional methods in models of with cluster like threshold influence patterns from environmental factors. These results highlight the potential of the proposed machine learning framework, with an accompanying R package currently under development to facilitate its application to complex datasets.

Introduction:Rotavirus remains a major cause of severe gastroenteritis among children under five years of age worldwide, including in Ethiopia. Despite the introduction of vaccines, the virus continues to evolve through frequent mutation and genetic reassortment, resulting in substantial genetic diversity and raising concerns about potential vaccine escape. This study aimed to assess the prevalence, genotype distribution, and genetic characteristics of rotavirus A (RVA) strains circulating among children with diarrhea in central Ethiopia, and to compare these strains with the Rotarix® vaccine strain.
Methods: A cross-sectional study was conducted between April 2022 and December 2023 in health centers located in Debre Birhan and Addis Ababa. Stool samples were collected from children under five years presenting with diarrhea. RVA detection was performed using quantitative real-time PCR (qPCR), and genotyping was carried out through Sanger sequencing of the VP7 and VP4 genes. Amino acid sequences were analyzed to identify substitutions within key antigenic epitopes in comparison with the Rotarix® vaccine strain.
Results: RVA was detected in 30 out of 247 samples (12.14%), with 28 successfully genotyped. G9 was the predominant G genotype (50%), followed by G12, G2, G1, and G3, while a proportion remained untyped. Among P genotypes, P[4] was most common, followed by P[6] and P[8]. The dominant G/P combination was G9P[4]. Notably, multiple amino acid substitutions were observed in both VP7 and VP4 antigenic regions compared to the Rotarix® strain. Conclusion: These findings suggest ongoing viral evolution and highlight the need for continuous surveillance and evaluation of vaccine effectiveness.

The South Caucasus region, including Armenia, is recognized as a center of early viticulture, the oldest known winery, and a tradition of winemaking. Armenia's topography has contributed to the preservation of genetically diverse grapevine populations. Cultivated grapevines (Vitis vinifera ssp. vinifera) and their wild ancestor (V. vinifera ssp. sylvestris) exhibit high genetic diversity, making them valuable resources for understanding domestication, adaptation, and breeding. Despite Armenia's historical and economic importance, the genomic diversity of its wild and cultivated grapevines remains underexplored. We re-analyzed whole-genome sequencing data of 164 grapevine accessions from Armenia to characterize genomic diversity, population structure, and domestication history.
Our analysis uncovered genetic patterns partly unique to Armenia. Population structure analysis revealed a genetic separation between wild and cultivated groups and three distinct ancestral components within the cultivated gene pool, reflecting a west-to-east geographical gradient in Armenia. This genetic cline correlates with a shift in usage, from table to wine grapes, and a transition in berry skin color from white to black. Additionally, we identified four distinct subgroups within wild populations in Syunik, suggesting notable diversity. Evolutionary history analysis indicates that wild and cultivated lineages began to separate ~18.5k years ago, with divergence intensifying ~4k years ago under human cultivation. Comparative genomic scans for divergent selection identified genomic regions associated with domestication traits. GWAS uncovered candidate markers linked to agronomic traits, such as berry skin color and bunch density.
The results underscore the importance of conserving local grapevine diversity in Armenia, a historically significant and genetically rich viticultural region.

The prediction of homologous genes is fundamental for functional annotation transfer, evolutionary analysis, and comparative genomics. Orthology inference tools such as Proteinortho, OrthoFinder, SonicParanoid, and OrthoMCL distinguish orthologs and paralogs but typically overlook gene fusion and fission events, which also play a key role in protein evolution by reshaping domain architectures and enabling functional diversification.
We demonstrate that orthology data can be leveraged to detect fusion and fission events, extending its utility beyond conventional applications. To this end, we developed FuFisOr, a lightweight post-processing pipeline designed to identify such events from orthology inference outputs, using Proteinortho as an example. The pipeline detects candidate fission sets aligning to potential fusion proteins by exploiting reciprocal best hits, avoiding additional similarity searches. FuFisOr integrates seamlessly with Proteinortho, requires minimal pre-processing, and is adaptable to other tools.
FuFisOr was evaluated on three bacterial proteomes: Pseudomonas aeruginosa, Bacillus subtilis, and Escherichia coli, identifying 15 high-confidence fusion/fission events validated using InterPro domain annotations and AlphaFold2 structural models superimposed in UCSF Chimera. One representative case involves the riboflavin biosynthesis protein RibBA in B. subtilis, which contains both DHBP_synthase and GTP_cyclohydro2 domains; in E. coli, these exist separately as RibB and RibA, indicating a fission event.
These results highlight FuFisOr ability to detect biologically meaningful events and extend orthology-based analyses with a scalable, reproducible solution for studying protein evolution.

Spatial transcriptomics is becoming an important approach for studying cancer tissue architecture, heterogeneity, and microenvironment organization. However, publicly available cancer datasets remain highly heterogeneous in terms of preprocessing, metadata quality, format variability, which limits their usability and complicates large-scale, cross-study analyses. To address this, we are developing OncoSpat, a pan-cancer spatial transcriptomics atlas and interactive portal that currently integrates 1,229 spatial samples from manually curated public Visium and related in situ-capturing transcriptomics datasets across human and mouse tumors. OncoSpat is designed as a harmonized and continuously expandable atlas framework rather than a static collection. At this stage, the project focuses on building a standardized sample registry with unified metadata, source information, spatial asset availability, and sample readiness information. In parallel, we are establishing a harmonization workflow for consistent sample formatting, gene identifier standardization, and quality-aware spatial preprocessing. Planned analysis layers include spot-level quality control, spatial graph diagnostics, tissue fragmentation, hotspot-based region definition, and co-localization analysis, with the broader aim of identifying conserved spatial phenotypes and recurring niches across cancers. We are also developing a scalable Python-based benchmarking module to evaluate data transformation strategies for Visium spatial transcriptomics at cohort scale. The framework is further being organized as a modular workflow system with checkpointed analysis stages, explicit quality control review steps, and structured tracking of datasets and outputs. The first public release is planned for next year, with continued expansion through integration of newly published public cancer spatial datasets.

T cells play a crucial role in eliminating cancer cells and pathogen-infected cells, underscoring their significance in cancer immunotherapy. The functionality of T cells depends on the recognition of peptides presented by HLA (human leukocyte antigen) molecules via the T-cell receptor (TCR) on the cell surface. This study aims to elucidate the interaction dynamics between peptide-HLA complexes and TCRs, focusing on the effects of different binding modes, HLA alleles, and epitope mutations on TCR recognition.
We stimulated human primary T cells with various epitopes to characterize these interactions. Our findings reveal that the bulge binding mode of peptides induces significant changes in the TCR library. Additionally, changes in the HLA molecule lead to substantial shifts in TCR recognition patterns. The extent of TCR library alteration varies depending on the mutation's location within the epitope and the degree of amino acid difference from the wild-type sequence.
These results not only uncover fundamental principles governing peptide-HLA and TCR interactions but also provide insights into predicting cross-reactivity between different peptide-HLA molecules. Understanding these mechanisms enhances our ability to forecast TCR responses to various epitopes, which is crucial for designing effective cancer immunotherapies and vaccines. This study lays the groundwork for leveraging public TCR datasets to predict cross-reactivity, potentially reducing the reliance on repetitive functional assays.

Copy number aberrations (CNAs) are a major class of somatic alterations in cancer and provide clinically actionable signals through oncogene amplifications and tumor suppressor deletions. Whole-exome sequencing (WES) is widely adopted in translational and clinical genomics, yet CNA calling from WES remains difficult—especially for formalin-fixed paraffin-embedded (FFPE) material and tumor-only workflows—because capture and GC biases, variable mappability, degraded DNA, and unknown purity/ploidy can obscure true copy number states and inflate false positives, particularly for focal events. In this work, we review the principles and limitations of WES-based CNA inference and outline practical strategies for normalization, segmentation, and quality control in challenging sample contexts. We further explore how additional genomic features (e.g., depth and allelic imbalance summaries, segment-level noise metrics, and concordance across callers) can be combined with machine learning (ML) approaches to improve confidence estimation and prioritization of CNA calls. As initial results, we describe a framework for leveraging multi-caller ensembles and feature-driven scoring to better separate likely biological CNAs from technical artefacts, with an emphasis on enhancing detection and interpretation of focal deletions and other clinically relevant aberrations in tumor-only and FFPE WES data.

Understanding the evolving training needs of researchers in computational life sciences is essential to ensure that training remains relevant, impactful, and accessible. The SIB Training Group delivers over 60 training events annually to a diverse audience spanning academia, industry, and healthcare sectors. To guide the future development of our training portfolio, we conducted a survey to identify emerging scientific topics, preferred learning formats, and barriers to participation.

Responses were collected across career stages, and scientific domains. A core section of the survey focused on identifying training topics not currently covered by the SIB portfolio. Using Likert-scale ratings, respondents evaluated the relevance of prospective topics across multiple areas of computational life sciences.

A second section examined barriers to training participation, including limited funding, time constraints during the work week, prerequisite knowledge requirements, and geographical distance. Participants rated the extent to which each factor affected their ability to attend SIB courses, enabling quantification of obstacles.

Finally, the survey assessed preferences for learning formats, including multi-day intensive courses, project-based learning, self-paced e-learning, webinars, blended approaches, AI-assisted learning, multi-week programmes, flipped classrooms, and bring-your-own-data models. These insights support the development of flexible and scalable training offerings aligned with researchers' needs and constraints.

The results provide an overview of the Swiss life science training landscape and will directly inform the design of the 2027 SIB Training Programme. This poster presents the survey design, preliminary analyses, and implications for future training development, offering a model applicable to other bioinformatics training initiatives.

Bioinformatics relies on standardized workflows executed through engines such as Nextflow, Snakemake, or Galaxy, producing large, heterogeneous outputs across multiple runs. Yet researchers still lack a unified way to aggregate, explore, and share these results interactively, without writing custom code for each project.
We present Depictio, an open-source web platform that transforms bioinformatics workflow outputs into interactive, shareable dashboards. Depictio's command-line interface scans workflow output directories, automatically detects file structures through configurable patterns, and ingests tabular data into a scalable backend. Data from multiple workflow executions are aggregated into unified, versioned collections ready for exploration.
A central recent feature is Depictio's template system. Community-driven and peer-reviewed dashboard templates, currently developed for nf-core pipelines, define both the data ingestion logic and the dashboard layout for a given workflow. A single CLI command points Depictio at a pipeline's output directory and automatically generates a fully populated, interactive dashboard without manual configuration required.
Dashboards are organized into multi-tab interactive views, each composed from a rich component library including interactive components, figures, tables, metric cards, MultiQC plots, image galleries, and geospatial maps. Cross-component filtering, through lasso, box-select and table selection propagates in real time across linked components, enabling fluid and live exploratory analysis.
Depictio is cloud-ready, deployable via Docker Compose or Kubernetes, and available as a hosted application on SciLifeLab Serve (https://serve.scilifelab.se) through a collaboration with SciLifeLab Data Centre. Depictio is released under the MIT license at https://github.com/depictio/depictio.

Do you know that 90% of data is never accessed three months after storage? Do you know that a third of the 8,000 plant types present in Switzerland in the 1990s can no longer be traced? Do you know that the manufacture of devices often has a higher carbon impact than their usage?

We, the EcoImpact Group, tackle these challenges at SIB. We promote eco-friendly commuting through Bike2Work, measure our collective carbon footprint to identify areas for improvement, and raise awareness of biodiversity with BioBlitz events using the iNaturalist app. We participate in Digital Cleanup Day to reduce digital clutter and will host a Learn@Lunch seminar to share knowledge on planetary boundaries and global environmental limits. Join us, propose your ideas, and make an impact at SIB and beyond.

Non-Ribosomal Peptide Synthetases (NRPSs) are modular enzymatic assembly lines that synthesise Non-Ribosomal Peptides (NRPs), secondary metabolites often used as therapeutics. Within NRPSs, Adenylation (A) domains dictate substrate specificity, governing the chemical composition of the NRP. Computational efforts have traditionally framed A-domain specificity prediction as a supervised classification problem, producing discrete predictions over a closed substrate vocabulary. This is at odds with NRPS research, where novel specificities are routinely characterised, requiring output-layer extension. We propose AD-CLIP, the first retrieval-based approach to A-domain specificity, embedding A-domains and substrates in a shared chemical-protein space where compatibility is a continuous similarity score and substrate vocabulary remains open. We curate 4,250 A-domain sequences spanning 43 substrates. We use 16 specificity-determining residues, partially overlapping with canonical Stachelhaus residues, identified in our prior classification work. The dual encoder pairs an MLP for substrates with an A-domain module of per-residue MLPs, an E(3)-equivariant graph neural network (EGNN) over Boltz2-predicted Cα coordinates, and mean pooling. We evaluate AD-CLIP using recall and Normalised Discounted Cumulative Gain (NDCG) under two scenarios: zero-shot retrieval, and n-shot continual pretraining (n ∈ {1, …, 5}), where pretraining continues on incrementally added examples per held-out substrate. Zero-shot retrieval achieves mean recall@200 = 0.61 and NDCG@200 = 0.45. AD-CLIP adapts rapidly under minimal supervision: a single example raises NDCG@200 / recall@200 to 0.74 / 0.85, averaging to 0.97 / 0.91 with five while avoiding catastrophic forgetting. AD-CLIP is thus effective and extensible, accommodating new substrates at minimal cost.

Transformer models have become central to biological data analysis, yet most genomic workflows still incorporate linkage disequilibrium (LD) through explicit preprocessing, typically relying on pairwise correlation estimates that are computationally demanding and task-specific. Here, we introduce LD-Attention, a generalized LD-aware attention layer as a modular and reusable transformer component. By embedding LD directly within attention, the model learns locus-to-locus dependency structure implicitly during end-to-end training, removing the need for precomputed LD matrices. In this formulation, LD is not treated as an external feature but as an inductive bias encoded within the architecture, enabling attention to align with the correlation structure of genetic variation. The resulting framework is task-agnostic and readily extensible across LD-dependent analyses, including genotype imputation, genome-wide association studies, fine-mapping, haplotype phasing, and polygenic risk prediction. By abstracting LD awareness into a portable attention layer, LD-Attention provides a scalable foundation for integrating LD structure into transformer-based models for genetic data.

Continuous genomic surveillance has been central to monitoring SARS-CoV-2 evolution and supporting public health decisions. Public health authorities, including the World Health Organization (WHO), monitor lineage dynamics to inform variant designation, including Variants of Interest (VOI). In our previous work, we developed CoVerage (sarscoverage.org), a large-scale analytics platform that tracks lineage dynamics and enables early detection of rapidly increasing lineages as potential variants of concern through statistical analysis of prevalence changes and phylogeny-informed lineage structure. While CoVerage provides high-resolution insights into lineage frequency trends and antigenic evolution, it relies on sequencing data, which has recently declined in many regions.

This reduction in sequencing effort introduces uncertainty in estimating population-level prevalence of circulating lineages, limiting robustness of downstream interpretation for variant monitoring and designation. To address this gap, we present a Bayesian probabilistic framework that integrates sequencing data with epidemiological case counts to estimate population-level lineage-attributed, confirmed case numbers under limited sampling.

We model observed lineage counts as multinomial realizations of latent lineage proportions and place a symmetric Dirichlet prior over these proportions. Posterior uncertainty is propagated via Monte Carlo sampling, and sampled proportions are scaled by weekly reported case counts to derive mean lineage-specific case estimates and 95% highest posterior density intervals.

We implemented this method within CoVerage, providing lineage-specific, confirmed case estimates at global and regional levels, for example per country. This extension enables uncertainty-aware inference of lineage dynamics in low-coverage settings, supporting more robust and timely variant monitoring and designation.

Oligonucleotides are short, synthetic strands of DNA or RNA that have many application areas. Based on the intended application and experimental design, researchers have to customize the length, sequence composition, and thermodynamic properties of the designed oligos. While various tools exist that provide customized oligo sequences, they each use their own implementation of the basic pipeline steps. This makes it difficult to create new pipelines. We tackle this issue with our open-source Oligo Designer Toolsuite (ODT). ODT is a collection of modules that provide all basic oligo design functionalities with a common underlying data structure within a flexible Python framework. This framework allows a lightweight development of custom oligo design pipelines. We have successfully developed custom probe design pipelines for cycleHCR, Oligo-seq, SCRINSHOT, SeqFISH+ and MERFISH protocols with ODT. ODT provides a command line interface, which can be hard to use by scientists unfamiliar with computational tools. Therefore, we currently develop a web service that provides a graphical user interface (GUI). ODT-Cloud runs on Helmholtz infrastructure and allows users to easily set parameters, select target regions and download a table with ready-to-order oligos via a GUI. ODT-Cloud also handles the automatic preprocessing of reference genomes and stores the results of the pipeline runs. ODT is available at https://github.com/HelmholtzAI-Consultants-Munich/oligo-designer-toolsuite and ODT-Cloud will be available at https://odt.helmholtz-munich.de/

Endemic human coronaviruses (HCoV) cause seasonal peaks of respiratory disease. As infections with endemic coronaviruses are usually mild, symptoms-based surveillance may provide an incomplete characterisation of the circulation of these viruses. This study evaluates the efficacy of wastewater-based epidemiology as a tool to monitor the circulation of endemic human coronaviruses HCoV-229E, HCoV-NL63, and HCoV-HKU1 in Zurich, Switzerland.
A multiplex digital PCR assay was used to quantify viral RNA in wastewater samples collected between December 2021 and December 2024 in Zurich, Switzerland. Using EpiSewer, a Bayesian semi-mechanistic model, we estimated the effective reproduction number and characterized epidemic waves. The findings were compared against clinical data from the Swiss Sentinel reporting system (Sentinella).
Aggregated weekly flow-normalised viral concentrations correlated with the weekly number of clinical cases reported through the Sentinella surveillance system, although the overall reported number of weekly cases was low. Circulation generally peaked in the first quarter of the year, though peaks occurred later in the season compared to pre-pandemic trends observed in the literature. HCoV-NL63 showed increased activity in early 2023 and late 2024, while HCoV-229E and HCoV-HKU1 showed peaks in winter 2022, 2023, and 2024. HCoV-229E circulated even during periods of non-pharmaceutical interventions due to the SARS-CoV-2 pandemic.
In summary, we demonstrate that wastewater surveillance can offer a robust, population-level approach to tracking endemic coronaviruses independent of healthcare-seeking behavior and can provide critical insights into the evolving seasonality of respiratory viruses in the post-pandemic era.

Protein-ligand interactions play a central role in many biological processes ranging from cell signaling to disease treatment. However, traditional drug discovery struggles with the immense number of potential drug molecules and the limited availability of detailed protein-ligand complex data. This bottleneck hinders the development of highly selective drugs with strong binding affinity to their target proteins. Recent research has opened an exciting new avenue by demonstrating the potential to predict the most probable orientations and positions of chemical groups solely from the local backbone conformation and the identity of the interacting amino acid. This eliminates the need for scarce protein-ligand complex data, offering a significant advantage in drug design.
Our research builds upon this promising avenue and takes a significant step forward by extending the concept to the broader protein micro-environment. We propose formulating the identification of functional hotspots within these micro-environments as a semantic segmentation task. Our method harnesses a convolutional neural network (CNN) trained on a meticulously curated database, encompassing a diverse array of protein micro-environments surrounding various functional groups. To assess the efficacy of our approach, we evaluated the predicted hotspots based on their pharmacological relevance and compared them to existing protein-ligand complexes. This evaluation demonstrates the significant potential of our deep learning approach to expedite the drug discovery process by prioritizing promising target sites for further investigation.

The Biomedical Data Science Facility (BDSF) provides bioinformatics services to biomedical research groups in the Lemanic region (largely UNIGE, HUG, UNIL, CHUV and EPFL). Our platform handles a comprehensive range of omics data, including single-cell RNA-sequencing, ATAC-sequencing, spatial transcriptomics, and metabolomics. Beyond analysis, we guide researchers in technology selection, statistical method implementation and data interpretability and sharing.

In this poster, we showcase projects that illustrate how we help transform complex biological data into meaningful scientific discoveries. These projects highlight the computational methods and analyses that our facility provides to the local scientific community. For example, we develop custom visualization tools, such as the B2CViz R package for spatial gene expression, and build interactive Shiny web applications (e.g., BrainTIME), which allow researchers to explore their own datasets without requiring advanced coding skills. Additionally, in collaboration with researchers and clinicians, the facility implements machine learning models for biomarker discovery. Finally, we also leverage the richness of publicly available data, which allows researchers to translate large-scale raw data into interpretable biological discovery.

Overall, the BDSF acts as a collaborative partner in navigating the full lifecycle of bioinformatics research, from technology selection to result interpretation. In addition to data analysis, we also offer training, including open-source courses in R, statistics, and omics data analysis, alongside tailored support for Ph.D. and postdoctoral researchers. We invite you to visit our poster to explore our projects in depth, learn more about our services, or discuss specific bioinformatics challenges in your own research.

Earth is teeming with microbial life, pervading all layers of the planet – from the deep subsurface to the upper atmosphere. Although characterizing this diversity remains a formidable challenge, microbial ecosystems are being explored at an unprecedented pace, continually revealing new taxa and communities. MicrobeAtlas integrates these studies into a unified, global resource for comparisons across environments, hosts, and technologies.

Here, we report on the latest updates and feature additions to MicrobeAtlas. The database now connects more than 5.3 million microbial sequencing samples into a unified dataset, substantially expanding environmental coverage and doubling community-level diversity. Using an improved, scalable workflow, these communities were clustered into tens of thousands of community types at adjustable resolution, enabling a bird's-eye view on global microbial ecosystems. Large language models now support the metadata extraction pipeline, improving environmental annotations, generating detailed keyword summaries, and enabling quantitative comparisons through semantic embeddings. Integration of the mOTUs4 and proGenomes4 databases, alongside existing links to BacDive, provides deeper functional and genomic context for lineages in MicrobeAtlas.

Ongoing efforts on reference database extension, strain-level biogeographic analyses, and microbial eukaryote integration will provide an increasingly comprehensive understanding of our microbial planet.

Cancer genomes are shaped by evolutionary processes that couple mutagenesis, clonal selection, chromosomal instability, spatial growth and treatment response into structured genomic patterns, yet current benchmarking strategies largely ignore this evolutionary dependency. Here, we present SCOUT, a large-scale synthetic whole-genome sequencing resource of over 200 samples, designed for systematic benchmarking of tumour genomic analysis and evolutionary inference under controlled evolutionary ground truth. Unlike conventional task-specific simulations, SCOUT models tumour evolution as a latent generative process that simultaneously shapes mutations, copy-number alterations, variant allele frequencies, mutational signatures and clonal architectures. SCOUT recapitulates key features of solid and haematological malignancies, including driver mutations, chromosomal instability, intratumour heterogeneity, spatial sampling and treatment-associated evolutionary dynamics in tumour and matched-normal longitudinal and multi-region sequencing designs. Using SCOUT, we benchmarked widely used methods for somatic variant detection, copy-number analysis, mutational signature inference and tumour evolutionary reconstruction. Across analytical tasks, performance deteriorated in low-purity, highly subclonal and structurally complex tumours, while spatial sampling bias and hypermutation generated spurious evolutionary signals that confounded tumour interpretation across multiple inference layers. Evolutionary simulations further distinguished lineage-restricted genetic bottlenecks from multi-lineage resistance dynamics associated with tumour plasticity. Tumour purity consistently exerted a stronger effect on inference accuracy than sequencing depth. Together, our results establish evolutionary ground truth as a prerequisite for reproducible benchmarking and biologically interpretable analysis of cancer whole-genome sequencing data.

Environmental exposures play a critical role in shaping physical and mental health, yet integrating such data into biomedical research remains technically complex and fragmented. The EnvironMENTAL Climate, Urbanicity, Environment and Society (CLUES) framework is an open-source, end-to-end workflow for generating individual-level environmental exposure data. CLUES automates the selection and download of open-access geospatial datasets, standardises spatial and temporal formats, and map projections, and links resulting environmental variables to individual-level biomedical data, requiring no prior expertise in geospatial data. CLUES covers key environmental domains, including urban and natural space, climate and weather extremes, air pollution, and regional socioeconomic conditions. Designed for extensibility and cross-cohort applicability, it enables multidimensional exposure mapping across global settings and adheres to FAIR (Findability, Accessibility, Interoperability and Reusability) and privacy-compliant data protection principles. In this work, we present the CLUES framework and evaluate its scalability, computational performance, and reproducibility for large-scale biomedical research.

Hepatitis C virus (HCV) vaccine development is fundamentally constrained by extreme genetic and antigenic diversity, which complicates rational antigen selection and the design of antigens inducing potent and broad neutralizing responses. Here, we apply a quantitative geometric framework to analyze and predict cross-neutralizing antibody responses based on antigen positioning in antigenic space.

Using a previously established antigenic cartography of virus-antibody interactions, we consider a low-dimensional embedding in which distances reflect functional neutralization similarity. Within this space, we introduce the concept of an antigenic center and derive distance-based metrics to characterize candidate immunogens and their combinations. We further integrate neutralization breadth and potency into a composite functional score, the cross-neutralization index (CNI), enabling consistent ranking across heterogeneous vaccine platforms.

Applying this framework to diverse recombinant E2 immunogens, we show that some, but not all, antigens closer to the antigenic center yield stronger cross-neutralizing responses, possibly linking spatial position to functional outcome. Moreover, viral antigen features independent of antigen cartography may influence antigen immunogenicity. Notably, rationally selected minimal antigen combinations that effectively cover this space outperform larger multi-cluster antigen pools, demonstrating that coverage--not antigen count--governs performance.

Anchored to human benchmarks of vaccine efficacy and HCV humoral immune protection, the framework provides a bridge computational modeling and translational vaccine evaluation. Together, this work demonstrates the use of a geometry-based approach for HCV vaccine antigen design, which may facilitate optimization of immunogens.