Biology subjects

Binder, H.

Publications and source records attributed to Binder, H..

At least 19 recordsLinked to original sources

bioRxiv2026

Coding agents author interpretable single-cell embedding models from the literature

The single-cell literature catalogs cell states as validated marker-gene programs -- a sparse, compositional prior. Conventional embedding methods do not leverage this prior and learn cell-state structure de novo from the expression matrix, producing dense dimensions needing post-hoc interpretation and batch correction. Here we show coding agents can author single-cell embedding models directly from the literature. Given a scenario that focuses this literature lens on a chosen biological subdomain, the agent edits a structured Python template, curating named, literature-cited gene programs and composing them into axes, without a gene-set database, training, or sight of the data. Across mouse and human tissues these zero-shot embeddings are competitive in biological quality with conventional, foundation-model, and program-informed baselines, batch-robust by construction and reproducible across runs, complementing data-driven embeddings. Because each dimension is a named, cited gene program, the embedding is interpretable and auditable, and its composable axes can be steered into a developmental tree.

bioinformatics↗

bioRxiv2026

Synthetic-data augmented calibration for expert-informed rare disease models

Clinical data for rare diseases are sparse, noisy, and heterogeneous, complicating calibration of ordinary differential equation (ODE) models. Thus, we introduce a noise-robust calibration in latent space that combines expertderived ODEs with learned latent representations. Our approach leverages synthetic ODE trajectories, augmenting our scarce observations to train a model-specific autoencoder representation and imputer. During calibration, observed and ODE-generated trajectories are compared in latent space, and ODE parameters are updated by minimizing their latent distance. In a controlled ABCDE simulation model, the imputer outperformed a carry-forward baseline for moderate parameter shifts, parameter recovery remained stable under random missingness, calibration remained robust to additional noise variables despite reduced downstream identifiability, and distinct dynamics formed visually separable latent trajectories. On a custom developed ODE model for real Epidermolysis Bullosa patients, the calibrated phenomenological model reproduced patient-level trajectories from sparse observations. Thus, we conclude that our latent-space calibration approach supports rare-disease modeling.

bioinformatics↗

bioRxiv2026

Mapping spatial cell-cell communication programs by tailoring chains of cells for transformer neural networks

Recent advances in spatial transcriptomics and computational modeling enable the study of cellular interactions in situ. However, existing methods quantify ligand-receptor activity pairwise or between predefined cell groups, yielding overlapping signals and limited ability to summarize concurrent interactions into programs while localizing communication hotspots. We introduce scCChain, a transformer-based framework that integrates ligand-receptor activity into spatially resolved communication programs and localizes hotspots at spot and single-cell resolution. scCChain first derives candidate programs using structured dimensionality reduction. Subsequently, it samples programspecific communication chains by linking transcriptionally similar sender cells to candidate receivers via weighted random walks on a distance-informed cell graph, borrowing signal from similar neighbors. Transformer-based modeling then scores chains to prioritize communication programs and pinpoint hotspots across the tissue. Applied to human breast cancer spatial transcriptomics data at spot and single-cell resolution, scCChain supports both exploratory communication program discovery and targeted analysis of user-specified ligand-receptor pairs. In spot-level data, it prioritizes a tumor-associated program enriched for pro-angiogenic signaling that localizes to invasive regions. In imaging-based data, it highlights CXCL12-CXCR4 communication hotspots at cellular resolution. Here, we demonstrate that chain-based transformer modeling enables interpretable discovery and mapping of biological meaningful spatial communication programs within complex tissues.

bioinformatics↗

bioRxiv2025

mmContext: an open framework for multimodal contrastive learning of omics and text data

SummaryMultimodal approaches are increasingly leveraged for integrating omics data with textual biological knowledge. Yet there is still no accessible, standardized framework that enables systematic comparison of omics representations with different text encoders within a unified workflow. We present mmContext, a lightweight and extensible multimodal embedding framework built on top of the open-source Sentence Transformers library. The software allows researchers to train or apply models that jointly embed omics and text data using any numeric representation stored in an AnnData .obsm layer and any text encoder available in Hugging Face. mmContext supports integration of diverse biological text sources and provides pipelines for training, evaluation, and data preparation. We train and evaluate models for a RNA-Seq and text integration task, and demonstrate their utility through zero-shot classification of cell types and diseases across four independent datasets. By releasing all models, datasets, and tutorials openly, mmContext enables reproducible and accessible multimodal learning for omics-text integration. Availability and implementationPretrained checkpoints and full source code for our custom MMContextEncoder are available on Hugging Face huggingface.co/jo-mengr. The Python package github.com/mengerj/mmcontext provides the model implementation and training and evaluation scripts for custom training.

bioinformatics↗

bioRxiv2025

Adding layers of information to scRNA-seq data using pre-trained language models

Pre-trained language models promise to enrich single-cell analyses with contextual information from large biomedical text corpora, but it remains unclear how to optimally align this knowledge with quantitative scRNA-seq data. To address this, we construct text-based training datasets from both scRNA-seq data and biomedical literature targeted to the experimental setting at hand. We then fine-tune lightweight encoder-only biomedical language models to learn a shared, literature-enriched representation. Controlled evaluations across immune and developmental datasets show that this representation preserves cell identity while adding robust and interpretable contextual layers of functional, disease-associated, and developmental information to single-cell analysis workflows.

bioinformatics↗

bioRxiv2025

PEtab Select: specification standard and supporting software for automated model selection

A central question in mathematical modeling of biological systems is determining which processes are most relevant and how they can be described. There are often competing hypotheses, which yield different models. Model comparison requires parameter optimization and sampling methods. Yet, standards for the specification of model selection problems and the swift evaluation of a broad spectrum of approaches are not available. PEtab Select addresses this challenge by providing a concise, standardized specification of model selection and its associated calibration problems through a new file format standard and software package. The standard facilitates the compact representation of even very large model selection problems; in one example, billions of model alternatives. PEtab Select builds on the PEtab standard for the specification of parameter estimation problems, and enables the use of state-of-the-art modelling and calibration workflows utilizing COPASI, Data2Dynamics, PEtab.jl, and pyPESTO. PEtab Select supports common model selection criteria (e.g., Akaike and Bayesian information criteria) and can be easily extended to use others. To ensure flexibility, PEtab Select implements several model space exploration approaches, including basic brute-force, forward, and backward selection, and also advanced, flexible selection methods. PEtab Select introduces the first standardization of model selection tasks, filling a critical gap in existing computational pipelines. It constitutes an essential contribution to FAIR research software in systems biology by promoting interoperability and reusability in model selection. Author summaryModel selection is a crucial step in mathematical modeling, guiding the choice of components to include in a model. PEtab Select automates this process across diverse modeling frameworks and programming languages via (1) a new interoperable, language-agnostic standard for specifying large-scale model selection problems, and (2) a comprehensive software package that implements these selection methods. Developed through a community effort, PEtab Select has been integrated into multiple modeling frameworks and is accessible to users of COPASI, Julia, MATLAB, and Python.

systems biology↗

bioRxiv2025

Evaluating discrepancies in dimensionality reduction for time-series single-cell RNA-sequencing data

There are various dimensionality reduction techniques for visually inspecting dynamical patterns in time-series single-cell RNA-sequencing (scRNA-seq) data. However, the lack of one-to-one correspondence between cells across time points makes it difficult to uniquely uncover temporal structure in a low-dimensional manifold. The use of different techniques may thus lead to discrepancies in the representation of dynamical patterns. However, The extent of these discrepancies remains unclear. To investigate this, we propose an approach for reasoning about such discrepancies based on synthetic time-series scRNA-seq data generated by variational autoencoders. The synthetic dynamical patterns induced in a low-dimensional manifold reflect biologically plausible temporal patterns, such as dividing cell clusters during a differentiation process. We consider manifolds from different dimensionality reduction techniques, such as principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), uniform manifold approximation and projection (UMAP) and single-cell variational inference (scVI). We illustrate how the proposed approach allows for reasoning about to what extent low-dimensional manifolds, obtained from different techniques, can capture different dynamical patterns. The results indicate that these techniques may not reliably represent dynamics when used in isolation. Thus, the proposed synthetic dynamical pattern approach provides a foundation for guiding future methods development to detect complex patterns in time-series scRNA-seq data.

bioinformatics↗

bioRxiv2024

Sparse dimensionality reduction for analyzing single-cell-resolved interactions

SummarySeveral approaches have been proposed to reconstruct interactions between groups of cells or individual cells from single-cell transcriptomics data, leveraging prior information about known ligand-receptor interactions. To enhance downstream analyses, we present an end-to-end dimensionality reduction workflow, specifically tailored for single-cell cell-cell interaction data. In particular, we demonstrate that sparse dimensionality reduction can pinpoint specific ligand-receptor interactions in relation to clusters of cell pairs. For sparse dimensionality reduction, we focus on the Boosting Autoencoder approach (BAE). Overall, we provide a comprehensive workflow, including result visualization, that simplifies the analysis of interaction patterns in cell pairs. This is supported by a Jupyter notebook that can readily be adapted to different datasets. Availability and implementationhttps://github.com/NiklasBrunn/Sparse-dimension-reduction Contactniklas.brunn@uniklinik-freiburg.de Supplementary material...

bioinformatics↗

bioRxiv2024

Enhancement of network architecture alignment in comparative single-cell studies

1Animal data can provide meaningful context for human gene expression at the single-cell level. This context can improve cell-type or cell-state detection and clarify how well the animal models human biological processes. To achieve this, we propose a deep learning approach that identifies a unified latent space to map complex patterns between datasets. Specifically, we combine variational autoencoders with a data-level nearest neighbor search to align neural network architectures across species. We visualize commonalities by mapping cell samples into the latent space. The aligned latent representation facilitates information transfer in applications of liver, white adipose tissue, and glioblastoma cells from various animal models. We also identify genes that exhibit systematic differences and commonalities between species. The results are robust for small datasets and with large differences in the observed gene sets. Thus, we reliably uncover and exploit similarities between species to provide context for human single-cell data.

bioinformatics↗

bioRxiv2024

Infusing structural assumptions into dimension reduction for single-cell RNA sequencing data to identify small gene sets

Dimensionality reduction greatly facilitates the exploration of cellular heterogeneity in single-cell RNA sequencing data. While most of such approaches are data-driven, it can be useful to incorporate biologically plausible assumptions about the underlying structure or the experimental design. We propose the boosting autoencoder (BAE) approach, which combines the advantages of unsupervised deep learning for dimensionality reduction and boosting for formalizing assumptions. Specifically, our approach selects small sets of genes that explain latent dimensions. As illustrative applications, we explore the diversity of neural cell identities and temporal patterns of embryonic development.

bioinformatics↗

bioRxiv2022

Designing Single-Cell RNA-Sequencing Experiments for Learning Latent Representations

To investigate the complexity arising from single-cell RNA-sequencing (scRNA-seq) data, researchers increasingly resort to deep generative models, specifically variational autoencoders (VAEs), which are trained by variational inference techniques. Similar to other dimension reduction approaches, this allows encoding the inherent biological signals of gene expression data, such as pathways or gene programs, into lower-dimensional latent representations. However, the number of cells necessary to adequately uncover such latent representations is often unknown. Therefore, we propose a single-cell variational inference approach for designing experiments (scVIDE) to determine statistical power for detecting cell group structure in a lower-dimensional representation. The approach is based on a test statistic that quantifies the contribution of every single cell to the latent representation. Using a smaller scRNA-seq data set as a starting point, we generate synthetic data sets of various sizes from a fitted VAE. Employing a permutation technique for obtaining a null distribution of the test statistic, we subsequently determine the statistical power for various numbers of cells, thus guiding experimental design. We illustrate with several data sets from various sequencing protocols how researchers can use scVIDE to determine the statistical power for cell group detection within their own scRNA-seq studies. We also consider the setting of transcriptomics studies with large numbers of cells, where scVIDE can be used to determine the statistical power for sub-clustering. For this purpose, we use data from the human KPMP Kidney Cell Atlas and evaluate the power for sub-clustering of the epithelial cells contained therein. To make our approach readily accessible, we provide a comprehensive Jupyter notebook at https://github.com/MTreppner/scVIDE.jl that researchers can use to design their own experiments based on scVIDE.

bioinformatics↗

bioRxiv2022

The performance of deep generative models for learning joint embeddings of single-cell multi-omics data

Recent extensions of single-cell studies to multiple data modalities raise new questions regarding experimental design. For example, the challenge of sparsity in single-omics data might be partly resolved by compensating for missing information across modalities. In particular, deep learning approaches, such as deep generative models (DGMs), can potentially uncover complex patterns via a joint embedding. Yet, this also raises the question of sample size requirements for identifying such patterns from single-cell multi-omics data. Here, we empirically examine the quality of DGM-based integrations for varying sample sizes. We first review the existing literature and give a short overview of deep learning methods for multi-omics integration. Next, we consider eight popular tools in more detail and examine their robustness to different cell numbers, covering two of the most common multi-omics types currently favored. Specifically, we use data featuring simultaneous gene expression measurements at the RNA level and protein abundance measurements for cell surface proteins (CITE-seq), as well as data where chromatin accessibility and RNA expression are measured in thousands of cells (10x Multiome). We examine the ability of the methods to learn joint embeddings based on biological and technical metrics. Finally, we provide recommendations for the design of multi-omics experiments and discuss potential future developments.

bioinformatics↗

bioRxiv2022

DOT1L activity affects cell lineage progression in the developing brain by controlling metabolic programs

Cortical neurogenesis depends on the tight balance between self-renewal and differentiation of apical progenitors (APs), the key progenitor type generating all other neural cells including neocortical neurons. We here report the activity of the histone methyltransferase DOT1L as a gatekeeper for AP cell identity. Combining lineage tracing with single-cell RNA sequencing of clonally related cells, we explore consequences of DOT1L inhibition on AP lineage progression during neurogenesis in the embryonic mouse neocortex. At the cellular level, DOT1L inhibition led to increased neurogenesis driven by a shift from asymmetric self-renewing to symmetric neurogenic divisions of APs. At the molecular level, we show that DOT1L activity preserved AP identity by promoting transcription of a gene set involved in AP metabolism. On a mechanistic level, DOT1L inhibition increased expression of metabolic genes, including microcephaly-associated Asparagine synthetase (Asns) and overexpression of ASNS in APs resulted in increased neuronal differentiation. Asns expression was predicted to be controlled through EZH2 and we show that DOT1L activity allows PRC2-mediated repression of Asns expression. Importantly, inhibition of ASNS activity rescued increased AP differentiation upon DOT1L inhibition. Our data show that DOT1L activity/PRC2 crosstalk controls AP lineage progression by regulating AP metabolism, and they provide a mechanistic view on how DOT1L activity might affect neocortical neurogenesis.

developmental biology↗

bioRxiv2021

Single-cell transcriptomic resolution of stem cells and their developmental trajectories in the hippocampus reveals epigenetic control of cell state perseverance

Despite conceptual research on hippocampus development and the application of single-cell-resolved technologies, the nature and maturation of its diverse progenitor populations are unexplored. The chromatin modifier DOT1L balances progenitor proliferation and differentiation, and conditional loss-of-function mice featured impaired hippocampus development. We applied single-cell RNA sequencing on DOT1L-mutant mice and explored cell trajectories in the E16.5 hippocampus. We resolved in our data five distinct neural stem cell populations with the developmental repertoire to specifically generate the cornu ammonis (CA) 1 field and the dentate gyrus (DG). Within the two developing CA1- and CA3-fields, we identified two distinct maturation states and we thus propose CA1- and CA3-differentiation along the radial axis. In the developing hippocampus, DOT1L is primarily involved in the proper development of CA3 and the DG, and it serves as a state-preserving epigenetic factor that orchestrates the expression of several important transcription factors that impact neuronal differentiation and maturation. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/452775v1_ufig1.gif" ALT="Figure 1"> View larger version (70K): org.highwire.dtl.DTLVardef@d611caorg.highwire.dtl.DTLVardef@2a0e3corg.highwire.dtl.DTLVardef@10c9082org.highwire.dtl.DTLVardef@1e358c8_HPS_FORMAT_FIGEXP M_FIG C_FIG O_LIThe developing hippocampus contains distinct and spatially separated NSC populations that differ in expression of a specific set of firstly described marker genes. C_LIO_LICA pyramidal neurons mature along the radial axis and pass through distinct maturation states. C_LIO_LIDOT1L preserves the dentate granule cell lineage in the developing hippocampus and limits maturation in the CA1- and CA3-fields development. C_LIO_LIDOT1L gates cell maturation as upstream regulator of transcription factor expression that confer instrumental roles in hippocampus development. C_LI

neuroscience↗

bioRxiv2021

Incorporating structural knowledge into unsupervised deep learning for two-photon imaging data

Live imaging techniques, such as two-photon imaging, promise novel insights into cellular activity patterns at a high spatio-temporal resolution. While current deep learning approaches typically focus on specific supervised tasks in the analysis of such data, we investigate how structural knowledge can be incorporated into an unsupervised generative deep learning model directly at the level of the video frames. We exemplify the proposed approach with two-photon imaging data from hippocampal CA1 neurons in mice, where we account for spatial structure with convolutional neural network components, disentangle the neural activity of interest from the neuropil background signal with separate foreground and background encoders and model gradual temporal changes by imposing smoothness constraints. Taken together, our results illustrate how such architecture choices facilitate a modeling approach that combines the flexibility of deep learning with the benefits of domain knowledge, providing an interpretable, purely image-based model of activity signals from live imaging data. Teaser sentenceUsing a neural network architecture that reflects domain knowledge provides an interpretable model of live cell imaging data.

bioinformatics↗

bioRxiv2021

Telomere maintenance pathway activity analysis enables tissue- and gene-level inferences

Telomere maintenance is one of the mechanisms ensuring indefinite divisions of cancer and stem cells. Good understanding of telomere maintenance mechanisms (TMM) is important for studying cancers and designing therapies. However, molecular factors triggering selective activation of either the telomerase dependent (TEL) or the alternative lengthening of telomeres (ALT) pathway are poorly understood. In addition, more accurate and easy-to-use methodologies are required for TMM phenotyping. In this study, we have performed literature based reconstruction of signaling pathways for the ALT and TEL TMMs. Gene expression data were used for computational assessment of TMM pathway activities and compared with experimental assays for TEL and ALT. Explicit consideration of pathway topology makes bioinformatics analysis more informative compared to computational methods based on simple summary measures of gene expression. Application to healthy human tissues showed high ALT and TEL pathway activities in testis, and identified genes and pathways that may trigger TMM activation. Our approach offers a novel option for systematic investigation of TMM activation patterns across cancers and healthy tissues for dissecting pathway-based molecular markers with diagnostic impact.

bioinformatics↗

bioRxiv2021

Stratified neural networks in a time-to-event setting

Deep neural networks are now frequently employed to predict survival conditional on omics-type biomarkers, e.g. by employing the partial likelihood of Cox proportional hazards model as loss function. Due to the generally limited number of observations in clinical studies, combining different data-sets has been proposed to improve learning of network parameters. However, if baseline hazards differ between the studies, the assumptions of Cox proportional hazards model are violated. Based on high dimensional transcriptome profiles from different tumor entities, we demonstrate how using a stratified partial likelihood as loss function allows for accounting for the different baseline hazards in a deep learning framework. Additionally, we compare the partial likelihood with the ranking loss, which is frequently employed as loss function in machine learning approaches due to its seemingly simplicity. Using RNA-seq data from the Cancer Genome Atlas (TCGA) we show that use of stratified loss functions leads to an overall better discriminatory power and lower prediction error compared to their nonstratified counterparts. We investigate which genes are identified to have the greatest marginal impact on prediction of survival when using different loss functions. We find that while similar genes are identified, in particular known prognostic genes receive higher importance from stratified loss functions. Taken together, pooling data from different sources for improved parameter learning of deep neural networks benefits largely from employing stratified loss functions that consider potentially varying baseline hazards. For easy application, we provide PyTorch code for stratified loss functions and an explanatory Jupyter notebook in a GitHub repository.

bioinformatics↗

bioRxiv2020

Synthetic observations from deep generative models and binary omics data with limited sample size

Deep generative models can be trained to represent the joint distribution of data, such as measurements of single nucleotide polymorphisms (SNPs) from several individuals. Subsequently, synthetic observations are obtained by drawing from this distribution. This has been shown to be useful for several tasks, such as removal of noise, imputation, for better understanding underlying patterns, or even exchanging data under privacy constraints. Yet, it is still unclear how well these approaches work with limited sample size. We investigate such settings specifically for binary data, e.g., as relevant when considering SNP measurements, and evaluate three frequently employed generative modeling approaches, variational autoencoders (VAEs), deep Boltzmann machines (DBMs) and generative adversarial networks (GANs). This includes conditional approaches, such as when considering gene expression conditional on SNPs. Recovery of pair-wise odds ratios is considered as a primary performance criterion. For simulated as well as real SNP data, we observe that DBMs generally can recover structure for up to 100 variables with as little as 500 observations, with a tendency of over-estimating odds ratios when not carefully tuned. VAEs generally get the direction and relative strength of pairwise relations right, yet with considerable under-estimation of odds ratios. GANs provide stable results only with larger sample sizes and strong pair-wise relations in the data. Taken together, DBMs and VAEs (in contrast to GANs) appear to be well suited for binary omics data, even at rather small sample sizes. This opens the way for many potential applications where synthetic observations from omics data might be useful.

bioinformatics↗