Biology subjects

Lobentanzer, S.

Publications and source records attributed to Lobentanzer, S..

8 recordsLinked to original sources

bioRxiv2026

Turning Domain Expertise into Multi-Dimensional Evaluation of Biomedical AI with Karenina

Language models and agents are increasingly used in biomedicine, but current benchmarks reward correct answers even when the underlying reasoning is flawed. Here we introduce Karenina, an open-source framework that turns expert knowledge into multi-dimensional evaluations of questions, conversations and autonomous agents. Illustrated in Question-Answer pairs, multi-turn conversations and autonomous data-analysis, these dimensions together moves evaluation beyond scoring, enabling trustworthy decision-making with AI in biomedicine.

bioinformatics↗

bioRxiv2026

CROssBARv2: A Unified Computational Framework for Heterogeneous Biomedical Data Representation and LLM-Driven Exploration

Biomedical discovery is hindered by fragmented, modality-specific repositories and uneven metadata, limiting integrative analysis, accessibility, and reproducibility. To address these challenges, we present CROssBARv2, a provenance-rich biomedical data-and-knowledge integration platform that unifies heterogeneous sources into a maintainable, scalable system. By consolidating diverse data types into an extensive knowledge graph enriched with standardised ontologies, rich metadata, and deep learning-based vector embeddings, CROssBARv2 alleviates the need for researchers to navigate multiple siloed databases and can facilitate downstream tasks, including predictive modelling and mechanistic reasoning, enabling applications such as drug repurposing and protein function prediction. The platform offers interactive graph exploration and embedding-based semantic search with CROssBAR-LLM, an intuitive natural language question-answering system that grounds large language model (LLM) outputs in the underlying knowledge graph to mitigate hallucinations. We assess CROssBARv2 through (i) multiple use-case analyses to test biological coherence and relational validity; (ii) knowledge-augmented biomedical question-answering benchmarks comparing CROssBAR-LLM against generalist LLMs; and (iii) a deep learning-based predictive modelling experiment for protein function prediction leveraging the heterogeneous structure of CROssBARv2. Collectively, CROssBARv2 provides a scalable, AI-ready, and user-friendly foundation that facilitates hypothesis generation, knowledge discovery, and translational research.

bioinformatics↗

bioRxiv2026

Towards Useful and Private Synthetic Omics: Community Benchmarking of Generative Models for Transcriptomics Data

BackgroundThe synthesis of anonymized data derived from real-world cohorts offers a promising strategy for regulatory-compliant and privacy-preserving biological data sharing, potentially facilitating model development that can improve predictive performance. However, the extent to which generative models can preserve biological signals while remaining resilient to adversarial privacy attacks in high-dimensional omics contexts remains underexplored. To address this gap, the CAMDA 2025 Health Privacy Challenge launched a community-driven effort to systematically benchmark synthetic and privacy-preserving data generation for bulk RNA-seq cohorts. ResultsBuilding on this initiative, we systematically benchmarked 11 generative methods across two cancer cohorts ([~]1,000 and [~]5,000 patients) over 978 landmark genes. Methods were evaluated across complementary axes of distributional fidelity, downstream utility, biological plausibility and empirical privacy risk, with emphasis on trade-offs between vulnerability to membership inference attacks (MIA) and other evaluation dimensions. Expressive deep generative models achieved strong predictive utility and differential expression recovery, but were often more vulnerable to membership inference risk. Differentially private methods improved resistance to attacks at the cost of reduced utility, while simpler statistical approaches offered competitive utility with moderate privacy risk and fast training. ConclusionsSynthetic bulk RNA-seq quality is inherently multi-dimensional and shaped by trade-offs between utility, biological preservation and privacy. Our results indicate that differences in model architecture drive distinct trade-offs across these axes, suggesting that model choice should align with dataset characteristics, intended downstream use and privacy requirements. Privacy risk should also be assessed using multiple complementary attack methods and, where possible, formal differential privacy protection.

bioinformatics↗

bioRxiv2025

Community-based biomedical context to unlock agentic systems

Large language models (LLMs) face reliability challenges stemming from hallucinations and insufficient access to validated scientific resources. Existing solutions are often fragmented and limited to specific applications, hindering broader adoption and interoperability. Here, we present Biomedical Context for Artificial Intelligence (BioContextAI), an open-source initiative centered on Model Context Protocol (MCP) servers to address these limitations. BioContextAI provides a community-oriented registry for discovering domain-specific MCP servers and a proof-of-concept server implementation that integrates widely-used biomedical knowledgebases. By enabling standardized access to validated scientific knowledge, BioContextAI aims to facilitate the development of composable agentic systems for biomedical research. Together, this work contributes to an emerging ecosystem of community-driven approaches for expanding the capabilities and reliability of biomedical AI systems.

systems biology↗

bioRxiv2024

Single Cell Integration Characterises Metaplasia in Inflammatory Intestinal Diseases

The gastrointestinal (GI) tract consists of connected organs, from the oral cavity to rectum, which function to ensure efficient nutrient uptake and barrier immunity. Diseases of the GI tract affect millions worldwide and as such there are now over 25 published single cell RNA-sequencing (scRNAseq) datasets surveying the GI tract, profiling specific anatomical regions, cell lineages, ages and diseases. To consolidate these efforts, we harmonised and integrated scRNAseq datasets across the whole GI tract from developing and adult human tissues, as well as newly generated data from preterm gut. We uniformly processed 385 samples from 189 healthy controls using a newly developed automated QC approach (scAutoQC). In total, our healthy reference contains [~]1.1 million cells which we annotated to a total of 137 fine-grained cell states. We anchor 13 published and 1 unpublished GI disease datasets covering gastric and colorectal (CRC) cancers, celiac disease, ulcerative colitis (UC) and Crohns disease (CD) to this reference, taking our atlas to a total of 1.6 million cells. We provide our atlas as a valuable resource to the community (available at gutcellatlas.org). Using this resource, we discover epithelial cell metaplasia arising from stem cells across intestinal inflammatory diseases (celiac, UC and CD) and CRC with transcriptional similarity to cells of the gastric and Brunners glands. Whilst previously linked to mucosal healing, we now implicate these cells in inflammation through recruitment of immune cells including T cells and neutrophils, and through direct interactions with T cells. Overall, we discover a shift in paradigm whereby changes in stem cells during inflammation lead to altered mucosal tissue architecture, which in turn contributes to ongoing inflammation. These findings highlight that in addition to barrier function, epithelial cells actively contribute to progression of inflammation which may be a function applicable to other tissues and diseases.

genomics↗

bioRxiv2023

MetalinksDB: a flexible and contextualizable resource of metabolite-protein interactions

Interactions between proteins and metabolites are key for cellular function, from the catalytic breakdown of nutrients to signaling. An important case is cell-cell communication, where cellular metabolites are secreted into the microenvironment and initiate a signaling cascade by binding to an intra- or extracellular receptor of another cell. While protein-protein mediated cell-cell communication is routinely inferred from transcriptomic data, for metabolite-protein interactions this is challenging due to the limitations of high-throughput single-cell and spatial metabolomics technologies, together with the absence of comprehensive prior knowledge resources that include metabolites. Here we report MetalinksDB, a comprehensive and flexible database of intercellular metabolite-protein interactions that is a magnitude larger than existing ones. MetalinksDB can be tailored to specific biological contexts such as diseases, pathways, or tissue/cellular locations by querying subsets of interactions using the web interface (https://metalinks.omnipathdb.org/) or the knowledge graph adapters. We showcase the use of MetalinksDB by identifying deregulated processes in renal cancer patients from multi-omics data as well as inferring metabolite-mediated cell-cell communication events driving acute kidney injury from spatial transcriptomic data. We anticipate that MetalinksDB will facilitate the study of metabolite-mediated communication processes. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=113 SRC="FIGDIR/small/573715v1_ufig1.gif" ALT="Figure 1"> View larger version (34K): org.highwire.dtl.DTLVardef@18c78aaorg.highwire.dtl.DTLVardef@145f21borg.highwire.dtl.DTLVardef@119d16borg.highwire.dtl.DTLVardef@3febb_HPS_FORMAT_FIGEXP M_FIG C_FIG

systems biology↗

bioRxiv2023

LIANA+: an all-in-one cell-cell communication framework

The growing availability of single-cell and spatially-resolved transcriptomics has led to the rapidly growing popularity of methods to infer cell-cell communication. Many approaches have emerged, each capturing only a partial view of the complex landscape of cell-cell communication. Here, we present LIANA+, a scalable framework to decode coordinated inter- and intracellular signalling events from single- and multi-condition datasets in both single-cell and spatially-resolved data. Beyond integrating and extending established methodologies and a rich knowledge base, LIANA+ enables novel analyses using diverse molecular mediators, including those measured in multi-omics data. Accessible as an open-source Python package at https://github.com/saezlab/liana-py, LIANA+ provides a comprehensive set of synergistic components to study cell-cell communication. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=124 SRC="FIGDIR/small/553863v1_ufig1.gif" ALT="Figure 1"> View larger version (53K): org.highwire.dtl.DTLVardef@16486ceorg.highwire.dtl.DTLVardef@ce6821org.highwire.dtl.DTLVardef@1165e54org.highwire.dtl.DTLVardef@18add5a_HPS_FORMAT_FIGEXP M_FIG C_FIG

systems biology↗

bioRxiv2019

Sex-related perturbations in schizophrenia and bipolar disorder brains reflect microRNA-mediated cholinergic/neurokine interactions

RNA-sequencing analyses are often limited to identifying lowest p-value transcripts, which does not address polygenic phenomena. To overcome this limitation, we developed an integrative approach that combines large scale transcriptomic meta-analysis of patient brain tissues with single-cell sequencing data of CNS neurons, short RNA-sequencing of human male- and female-originated cell lines, and connectomics of transcription factor- and microRNA-interactions with perturbed transcripts. We used this pipeline to analyze cortical transcripts of schizophrenia and bipolar disorder patients. While these pathologies show massive transcriptional parallels, their clinically well-known sexual dimorphisms remain unexplained. Our method explicates the differences between afflicted men and women, and identifies disease-affected pathways of cholinergic transmission and gp130-family neurokine controllers of immune function, interlinked by microRNAs. This approach may open new perspectives for seeking biomarkers and therapeutic targets, also in other transmitter systems and diseases.

molecular biology↗