Biology subjects

Zupan, B.

Publications and source records attributed to Zupan, B..

7 recordsLinked to original sources

bioRxiv2026

Single-cell foundation models transfer across deep evolutionary distances

Single-cell foundation models are trained mostly on human and other vertebrate cells, yet are proposed as general-purpose models. Whether they transfer far beyond their training species is unknown. We evaluated zero-shot embeddings from Geneformer, scGPT and UCE against PCA and scVI in developmental atlases of Dictyostelium discoideum and Caenorhabditis elegans, separated from humans by 1.2 billion and 600 million years. Without retraining, the models recovered developmental order and preserved organism-specific cell identities. UCE encodes genes by protein sequence and could use 97% of measured genes, whereas Geneformer and scGPT, after ortholog mapping, used under one-fifth. UCE performed most consistently, often exceeding species-native baselines. Restricted to ortholog-mappable genes, it still scored above vocabulary-based models on five of six annotation tasks, some narrowly. Foundation models can thus yield informative cell embeddings across more than a billion years of evolution, and protein-sequence gene representations make these embeddings more complete and often more accurate.

bioinformatics↗

bioRxiv2026

dictyExpress: an integrated browser for bulk and single-cell Dictyostelium transcriptomics

Summary dictyExpress 3.0 is a React/TypeScript reimplementation of the Dictyostelium discoideum transcriptomics web server, first introduced in 2009, that unifies bulk and single-cell exploration. A long-standing gene-expression resource for the Dictyostelium community, dictyExpress is also accessible through dictyBase, the central organism database for Dictyostelium. The new release preserves the curated bulk RNA-seq dashboards of earlier versions while adding an interactive module for a developmental scRNA-seq time course of wild-type AX4 and two cAMP-signalling mutants. Following visual analytics principles, interactive visualizations are coordinated through brushing and linking, with bulk time courses, differential expression, gene-ontology enrichment and the UMAP workspace sharing a common gene selection, enabling bookmarkable and exportable cross-modal analysis in one interface for Dictyostelium transcriptomics. Availability and implementation dictyExpress is available at https://app.dictyexpress.org/. Source code: https://github.com/biolab/dictyexpress-js. Contact lena.trnovec@fri-uni-lj.si, blaz.zupan@fri.uni-lj.si

bioinformatics↗

bioRxiv2025

Visual analytics framework for survival analysis and biomarker discovery

We introduce a visual analytics methodology for survival analysis, and propose a framework that defines a reusable set of visualization and modeling components to support exploratory and hypothesis-driven biomarker discovery. Survival analysis--essential in biomedicine--evaluates patients survival rates and the onset of medically relevant events, given their clinical and genetic profiles and genetic predispositions. Existing approaches often require programming expertise or rely on inflexible analysis pipelines, limiting their usability among biomedical researchers. The lack of advanced, user-friendly tools hinders problem solving, limits accessibility for biomedical researchers, and restricts interactive data exploration. Our methodology emphasizes functionality-driven design and modularity, akin to combining LEGO bricks to build tailored visual workflows. We (1) define a minimal set of reusable visualization and modeling components that support common survival analysis tasks, (2) implement interactive visualizations for discovering survival cohorts and their characteristic features, and (3) demonstrate integration within an existing visual analytics platform. We implemented the methodology as an open-source add-on to Orange Data Mining and validated it through use cases ranging from Kaplan-Meier estimation to biomarker discovery. The resulting framework illustrates how methodological design can drive intuitive, transparent, and effective survival analysis. Author summaryWhen studying diseases like cancer, it is important to understand how factors such as genes or treatments influence how long patients live after medical intervention. Survival analysis is a well-established statistical and bioinformatics approach for uncovering such insights, but it often requires advanced programming skills, which can be a barrier for many clinicians and life science researchers. To help overcome this, we developed a visual analytics tool that makes the analysis of censored data easier, more interactive, and accessible. Instead of writing code, researchers can use our framework to assemble analysis components into workflows through visual programming. They can explore their data by comparing survival curves, identifying meaningful patient subgroups, and discovering potential biomarkers. The tool is implemented as part of Orange Data Mining, a free and open-source data analytics platform. We tested it on real-world cancer datasets to demonstrate how easily and quickly researchers can prototype powerful data analysis workflows. Our goal is to make survival analysis more accessible, empowering researchers to generate and test hypotheses without needing to program. Our goal is to make survival analysis more accessible, empowering researchers to generate and test hypotheses without programming, and contributing to the broader democratization of data science.

bioinformatics↗

bioRxiv2024

An Infralimbic Cortex Neuronal Ensemble Encoded During Learning Attenuates Fear Generalization Expression

Generalization allows previous experience to adaptively guide behavior when conditions change. The infralimbic (IL) subregion of the ventral medial prefrontal cortex plays a known role in generalization processes, although mechanisms remain unclear. A basic physical unit of memory storage and expression in the brain are sparse, distributed groups of neurons known as ensembles (i.e., the engram). Here, we set out to determine whether neuronal ensembles established in the IL during learning contribute to generalized responses. Generalization was tested in male and female mice by presenting a novel, ambiguous, tone generalization stimulus following Pavlovian defensive (fear) conditioning. The first experiment was designed to test a role for IL in generalization using chemogenetic manipulations. Results show IL regulates defensive behavior in response to ambiguous stimuli. IL silencing led to a switch in defensive state, from vigilant scanning to generalized freezing, while IL stimulation reduced freezing in favor of scanning. Leveraging activity-dependent "tagging" technology (ArcCreERT2 x eYFP system), a neuronal ensemble, preferentially located in IL Layer 2/3, was associated with the generalization stimulus. Remarkably, in the identical discrete location, fewer reactivated neurons were associated with the generalization stimulus at the remote timepoint (30 days) following learning. When an IL neuronal ensemble established during learning was selectively chemogenetically silenced, generalization increased. Conversely, IL neuronal ensemble stimulation reduced generalization. Overall, these data identify a crucial role for IL in suppressing generalized responses. Further, an IL neuronal ensemble, formed during learning, functions to later attenuate the expression of generalization in the presence of ambiguous threat stimuli.

neuroscience↗

bioRxiv2021

Transcriptional milestones in Dictyostelium development

Development of the social amoeba Dictyostelium discoideum begins by starvation of single cells and ends in multicellular fruiting bodies 24 hours later. These major morphological changes are accompanied by sweeping gene expression changes, encompassing nearly half of the 13,000 genes in the genome. To explore the relationships between the transcriptome and developmental morphogenesis, we performed time-series RNA-sequencing analysis of the wild type and 20 mutant strains with altered morphogenesis. These strains exhibit arrest at different developmental stages, accelerated development, or terminal morphologies that are not typically seen in the wild type. Considering eight major morphological transitions, we identified 1,371 milestone genes whose expression changes sharply between two consecutive transitions. We also identified 1,099 genes as members of 21 regulons, which are groups of genes that remain coordinately regulated despite the genetic, temporal, and developmental perturbations in the dataset. The gene annotations in these milestones and regulons validate known transitions and reveal several new physiological and functional transitions during development. For example, we found that DNA replication genes are co-regulated with cell division genes, so they are co-expressed in mid-development even though chromosomal DNA is not replicated at that time. Altogether, the dataset includes 486 transcriptional profiles, across developmental and genetic conditions, that can be used to identify new relationships between gene expression and developmental processes and to improve gene annotations. We demonstrate the utility of this resource by showing that the cycles of aggregation and disaggregation observed in allorecognition-defective mutants involve a dedifferentiation process. We also show unexpected variability and sensitivity to genetic background and developmental conditions in two commonly used genes, act6 and act15, and robustness of the coaA gene. Finally, we propose that gpdA should be used as a standard for mRNA quantitation because it is less sensitive to genetic background and developmental conditions than commonly used standards. The dataset is available for democratized exploration without the need for programming skills through the web application dictyExpress and the data mining environment Orange.

genomics↗

bioRxiv2019

openTSNE: a modular Python library for t-SNE dimensionality reduction and embedding

SummaryPoint-based visualisations of large, multi-dimensional data from molecular biology can reveal meaningful clusters. One of the most popular techniques to construct such visualisations is t-distributed stochastic neighbor embedding (t-SNE), for which a number of extensions have recently been proposed to address issues of scalability and the quality of the resulting visualisations. We introduce openTSNE, a modular Python library that implements the core t-SNE algorithm and its extensions. The library is orders of magnitude faster than existing popular implementations, including those from scikit-learn. Unique to openTSNE is also the mapping of new data to existing embeddings, which can surprisingly assist in solving batch effects.\n\nAvailabilityopenTSNE is available at https://github.com/pavlin-policar/openTSNE.\n\nContactpavlin.policar@fri.uni-lj.si, blaz.zupan@fri.uni-lj.si

bioinformatics↗

bioRxiv2019

Embedding to Reference t-SNE Space Addresses Batch Effects in Single-Cell Classification

Dimensionality reduction techniques, such as t-SNE, can construct informative visualizations of high-dimensional data. When working with multiple data sets, a straightforward application of these methods often fails; instead of revealing underlying classes, the resulting visualizations expose data set-specific clusters. To circumvent these batch effects, we propose an embedding procedure that takes a t-SNE visualization constructed on a reference data set and uses it as a scaffold for embedding new data. The new, secondary data is embedded one data-point at the time. This prevents any interactions between instances in the secondary data and implicitly mitigates batch effects. We demonstrate the utility of this approach with an analysis of six recently published single-cell gene expression data sets containing up to tens of thousands of cells and thousands of genes. In these data sets, the batch effects are particularly strong as the data comes from different institutions and was obtained using different experimental protocols. The visualizations constructed by our proposed approach are cleared of batch effects, and the cells from secondary data sets correctly co-cluster with cells from the primary data sharing the same cell type.

bioinformatics↗