Biology subjects

Sprang, M.

Publications and source records attributed to Sprang, M..

12 recordsLinked to original sources

bioRxiv2026

Prot2Prop: Structure-informed multitask protein property prediction

Protein engineering often relies on separate models for related developability properties, limiting efficiency and transfer across tasks. We present Prot2Prop, a multitask framework based on a frozen ProstT5 encoder with shared and task-specific adapters for joint prediction of six protein properties: material production, solubility, temperature stability, aggregation propensity, expression yield, and folding stability. Across held-out test data, Prot2Prop achieved strong performance on both classification and regression tasks, including AUROC values ranging from 0.86 to 0.98 for classification endpoints and Spearman correlations ranging from 0.73 to 0.86 for regression endpoints. The model achieved particularly strong performance for temperature stability (AUROC = 0.98) and aggregation propensity (Spearman = 0.86). Post-hoc calibration further improved regression accuracy, reducing folding stability MAE from 0.67 to 0.48. These results demonstrate that parameter-efficient multitask adaptation of protein language models can provide accurate and unified prediction of diverse protein developability properties. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=132 SRC="FIGDIR/small/735009v1_ufig1.gif" ALT="Figure 1"> View larger version (51K): org.highwire.dtl.DTLVardef@904e5forg.highwire.dtl.DTLVardef@954cdorg.highwire.dtl.DTLVardef@9e9417org.highwire.dtl.DTLVardef@10c9118_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2026

Age-driven Dysregulation of murine Dendritic Cells is controlled by cell-intrinsic and extrinsic effects

Aging is associated with chronic, low-grade inflammation and progressive immune dysfunction. However, the current understanding of age-associated changes in dendritic cells across tissues is scarce. Studies exploring ageing-associated changes in dendritic cells (DCs) have reported either a general decline in the overall DC compartment or subset-specific alterations affecting cDC1, cDC2, and pDC populations across spleen, lung and liver, underscoring the considerable inconsistencies across tissues and studies. To underpin whether age-associated changes are extrinsic or intrinsic we investigated DCs across bone marrow, six peripheral tissues and in in vitro bone marrow derived DC cultures to examine the effects of aging on DC-poiesis, tissue distribution, and cellular states related to DC functionality and activation. We discovered that aging selectively alters DC development in the bone marrow by reducing cDC progenitor populations while preserving pDC-poiesis. In peripheral tissues, however, age-associated changes in DC homeostasis were strongly tissue-dependent. The most significant shifts in cDC1 and cDC2 frequencies occurred in barrier tissues, such as the lung and small intestine. In contrast, the spleen and liver exhibited more limited or variable changes. These quantitative alterations were accompanied by tissue-specific changes in phenotypic and activation-associated markers, including CD24, CD103, CD11b, MHCII, and CD86. Single-cell transcriptomic analyses of senescent p21-expressing DC across tissues and subsets indicated localized inflammatory states that aligned with local macrophage populations, pointing toward cell-extrinsic niches that contribute to local age-associated dysfunction. Notably, aged bone marrow retained the capacity to efficiently generate DCs in in vitro Flt3L cultures, and antigen-presenting function of BMDC to CD4 and CD8 T cells was maintained, pointing towards preserved cell-intrinsic functions, albeit subset-specific differences in activation and inhibitory receptor expression in response to different pattern-recognition receptor agonists. Collectively, our findings indicate that aging does not superimpose a uniform alteration module to the DC compartment across tissues, but instead promotes selective alterations in DC ontogeny and tissue-specific remodeling of DC phenotypes and cellular states.

immunology↗

bioRxiv2026

Cross-Domain Transfer Learning from Peptides to Lipids Using a Multi-Property Fine-Tuned LLM

Accurate knowledge of liquid chromatography retention time (RT) is essential for confident compound identification in metabolomics and lipidomics. Yet, it is often constrained by the scarcity of experimental data for many molecular classes. Current workflows depend on experimental RT libraries, which are time-consuming to build and limited to previously observed compounds. Here, we present a transfer learning pipeline that leverages large, publicly available peptide datasets to enable accurate lipid RT prediction in data-sparse scenarios. We first show that a ChemBERTa language model, when pre-trained on peptides with a multi-task objective (predicting both RT and fundamental RDKit molecular descriptors), learns a more robust and generalizable chemical representation than a single-task (RT-only) model. This multi-property pre-training yielded superior generalization in lipids, achieving test R2 values of 0.842 against 0.814 (RT-only). Crucially, transferring this peptide-based model to lipid data provided a pronounced advantage in data-sparse scenarios. When fine-tuned on only 5% of available lipid data, the transferred model improved the median test R2 by +0.234 over a model trained from scratch. Significant benefits persisted at intermediate data scales (50-75%), with performance converging only when 100% of the lipid data was used. Notably, the pre-trained model never underperformed the baseline, exhibiting more stable training across all data scales. These results demonstrate that multi-property pre-training guides language models towards chemically meaningful representations that support better RT prediction in different molecular domains. Furthermore, peptide-based pre-training facilitates cross-domain transfer of chemical properties to lipid. Our work provides a practical, scalable strategy to mitigate data scarcity in lipidomics by transferring knowledge from data-rich peptide databases, offering a computational alternative to extensive experimental library generation and enabling more confident identification in small-scale omics studies.

bioinformatics↗

bioRxiv2025

Target-site Dynamics and Alternative Polyadenylation Explain Large Share of Apparent MicroRNA Differential Expression

MicroRNA (miRNA) abundance reflects a dynamic balance between biogenesis, target engagement, and decay, yet differential expression analyses typically ignore changes in target-site availability driven by alternative polyadenylation (APA). We introduce MIRNAPEX, an expression-stratification-based machine learning framework that quantifies miRNA regulatory effect sizes from RNA-seq data by integrating target-gene expression with 3'UTR isoform usage to infer effective binding-site dosage. Using pan-cancer training sets, we train models that learn relationships between transcriptomic features and miRNA log-fold changes, with APA patterns providing predictive information beyond gene expression alone. When applied to knockdowns of core APA regulators, MIRNAPEX captured widespread 3'UTR shortening and accurately anticipated miRNA-specific shifts whose direction and magnitude mirrored APA-driven changes in binding-site availability. Analysis of target-directed miRNA degradation interactions further showed that loss of distal decay-trigger sites coincides with increased miRNA abundance, consistent with reduced degradation. Together, these findings demonstrate that apparent miRNA differential expression can arise from dynamic target-site landscapes rather than altered miRNA transcription, and that neglecting this dimension can lead to misestimation of regulatory effect sizes.

bioinformatics↗

bioRxiv2025

Depletion of CX3CR1+ macrophages results in disrupted functionality and immune surveillance within epididymis and testis

A finely tuned immune regulation within the epididymis and testis is essential for male reproductive health. This balance is especially critical in the epididymis, where sperm mature and ascending infections frequently disrupt homeostasis, resulting in regionally different immune responses and potential long-term fertility impairments. We previously demonstrated that the epididymis harbors a region-specific immunological scaffold, with CX3CR1+ macrophages as the most prominent epithelium-associated immune cell population. Here, we established a transgenic mouse model to selectively deplete these intraepithelial CX3CR1+ macrophages within the epididymis, resulting in focal epithelial damage and impaired sperm maturation processes essential for proper sperm functionality. Additionally, a mild reduction of the testicular macrophage pool resulted in transient disruptions in spermatogenesis and steroidogenesis. Although the macrophage niche was repopulated after depletion, the newly recruited cells displayed altered phenotypes consistent with persistent sperm alterations. Following infection with uropathogenic Escherichia coli (UPEC), macrophage-depleted mice exhibited exacerbated immune responses - particularly in normally protected proximal epididymal regions - with earlier onset and more severe tissue damage. Transcriptomic analysis revealed a failure to restrain inflammatory responses, especially in genes involved in immune regulation and antibacterial defense, accompanied by elevated immune cell infiltration in infected macrophage-depleted mice. Overall, our findings confirm a crucial role for CX3CR1 macrophages in preserving epithelial integrity and modulating immune responses, supporting a stable tissue environment necessary for efficient organ function of both epididymis and testis. Significance statementMaintaining immune balance in the epididymis is essential for tissue health and protection against ascending infections. Using a transgenic mouse model that allows for selective depletion of CX3CR1 macrophages, this study examines their role in both the epididymis and testis under normal and infectious conditions. The results show that the removal of these macrophages causes localized epithelial damage, changes in immune cell make-up, and increased inflammation in the epididymis after bacterial infection, while also causing mild, reversible problems with spermatogenesis and steroid production in the testis. These findings support the idea that CX3CR1 macrophages contribute to region-specific immune regulation and epithelial stability--key features for keeping the tissue environment suitable for proper sperm development.

immunology↗

bioRxiv2025

The quest continues: Human CD4+ CD16+ CD56+ "exTreg" resemble NKT cells instead

Regulatory T cells are essential for immune tolerance, but their loss of function under inflammatory conditions in murine models signify a risk factor for Treg-based therapies. Recently CD4+ CD56+ CD16+ T cells were suggested to resemble such ex-Treg in human PBMC. Here, we re-evaluate the identity of the CD4+ CD56+ CD16+ population at a phenotypic and transcriptomic level using multiparametric flow cytometry on human PBMC and CITE-seq analysis to demonstrate that the CD4+ CD56+ CD16+ cells mostly constitute NKT cells instead. Further, we evaluated the stability of human Treg under lineage-challenging conditions and observe robust lineage stability in vitro. Finally, we also explore the potential of Tr17 induction using TGF-{beta} and IL-6, a possible therapeutic strategy for Treg ex vivo expansion-based therapies. Together, we conclude that human exTreg remain to be described and instead human Treg present as remarkably stable, further promoting Treg-based adoptive transfer therapies.

immunology↗

bioRxiv2025

Integrating the ENCODE blocklist for machine learning quality control of ChIP-seq with seqQscorer

MotivationQuality assessment of next-generation sequencing data is a complex but important task to ensure correct conclusions from experiments in molecular biology, biomedicine, and biotechnology. We previously introduced seqQscorer, a quality assessment tool using machine learning to support this process. To improve seqQscorer in terms of accuracy and processing time, we integrated the ENCODE blocklist* to derive a new type of quality-related features, supposed to be more informative and faster in generation than those conventionally used by seqQscorer. ResultsThe novel seqQscorer extension, called seqBLQ, allows us to improve the quality assessment for ChIP-seq data derived from human tissues and cell lines. Furthermore, seqBLQ enhances the usability of the tool by simplifying the installation procedure and reducing the computational resources required for feature generation. Availability and implementationhttps://github.com/salbrec/seqQscorer

bioinformatics↗

bioRxiv2025

Evaluating Genetic Regulators of MicroRNAs Using Machine Learning Models

This study explores the genetic regulators of microRNAs (miRNAs) using an ensemble of machine learning models to predict miRNA expression levels from gene expression data. Employing ridge regression, we accurately predicted the expression of 353 human miRNAs (R2 > 0.5), revealing robust miRNA-gene regulatory relationships. By analyzing the coefficients of these predictive models, we identified genetic regulators for each miRNA and highlighted the multifactorial nature of miRNA regulation. Further network analysis uncovered that miRNAs with higher predictive accuracy are more densely connected to their top predictive genes, reflecting strong regulatory control within miRNA-gene networks. To refine these insights, we filtered the gene-miRNA interaction networks to identify miRNAs specifically associated with enriched pathways, such as synaptic function and cardiovascular processes. From this pathway-centric analysis, we present a curated list of miRNAs and their genetic regulators, pinpointing their activity within distinct biological contexts. Additionally, our study provides a comprehensive set of metrics and coefficients for the genes most predictive of miRNA expression, along with a filtered subnetwork of miRNAs linked to specific pathways and phenotypes. By integrating miRNA expression predictors with network analysis and pathway enrichment, this work advances our understanding of miRNA regulatory mechanisms and their roles across distinct biological systems.

bioinformatics↗

bioRxiv2025

Life as a Function: Why Transformer Architectures Struggle to Gain Genome-Level Foundational Capabilities

Recent advances in generative models for nucleotide sequences have shown promise, but their practical utility remains limited. In this study, we explore DNA as a complex functional representation of evolutionary processes and assess the ability of transformer-based models to capture this complexity. Through experiments with both synthetic and real DNA sequences, we demonstrate that current transformer architectures, particularly auto-regressive models relying on next-token prediction, struggle to effectively learn the underlying biological functions. Our findings suggest that these models face inherent limitations, that cannot be overcome with scale, highlighting the need for alternative approaches that incorporate evolutionary constraints and structural information. We propose potential future directions, including the integration of topological methods or the switch of modelling paradigms, to enhance the generation of genomic sequences.

synthetic biology↗

bioRxiv2024

Visible neural networks for multi-omics integration: a critical review

Biomarker discovery and drug response prediction is central to personalized medicine, driving demand for predictive models that also offer biological insights. Biologically informed neural networks (BINNs), also known as visible neural networks (VNNs), have recently emerged as a solution to this goal. BINNs or VNNs are neural networks whose inter-layer connections are constrained based on prior knowledge from gene on-tologies pathway databases. These sparse models enhance interpretability by embedding prior knowledge into their architecture, ideally reducing the space of learnable functions to those that are biologically meaningful. In this systematic review--the first of its kind-- we identify 86 recent papers implementing such models and highlight key trends in architectural design decisions, data sources and methods for evaluation. Growth in popularity of the approach is apparently mitigated by a lack of standardized terminology, tools and benchmarks.

bioinformatics↗

bioRxiv2023

Unveiling IRF4-steered regulation of context-dependent effector programs in Th17 and Treg cells

The transcription factor interferon regulatory factor 4 (IRF4) is crucial for the differentiation and fate determination of pro-inflammatory T helper (Th)17 and the functionally opposing group of immunomodulatory regulatory T (Treg) cells. However, molecular mechanisms of how IRF4 steers diverse transcriptional programs in Th17 and Treg cells are far from being definitive. To unveil IRF4-driven lineage determination in Th17 and Treg cells, we integrated data derived from affinity-purification and full mass spectrometry-based proteome analysis with chromatin immune precipitation sequencing (ChIP-Seq). This allowed the characterization of subtype-specific molecular programs and the identification of novel, previously unknown IRF4 interactors in the Th17/Treg context, such as ROR{gamma}t, AHR, IRF8, BACH2, SATB1, and FLI1. Moreover, our data reveal that most of these transcription factors are recruited to IRF composite elements for the regulation of cell type-specific transcriptional programs providing a valuable resource for studying IRF4-mediated gene regulatory programs in pro- and anti-inflammatory immune responses.

immunology↗

bioRxiv2022

An outstanding bacterial membranome requires CRISPR-Cas systems to avoid the intervention of phages

Antimicrobial resistance is widely recognized as a serious global public health problem. To combat this threat, a thorough understanding of bacterial genomes is necessary. The current wide availability of bacterial genomes provides us with an in-depth understanding of the great variability of dispensable genes and their relationship with antimicrobials. Some of these accessory genes are those involved in CRISPR-Cas systems, which are acquired immunity systems that are present in part of bacterial genomes. They prevent viral infections through small DNA fragments called spacers. But the vast majority of these spacers have not yet been associated with the virus they recognize, and this has been named CRISPR dark matter. By analyzing the spacers of tens of thousands of genomes from six bacterial species highly resistant to antibiotics, we have been able to reduce the CRISPR dark matter from 80-90% to as low as 15% in some of the species. In addition, we have observed that, when a genome presents CRISPR-Cas systems, this is accompanied by particular collections of membrane proteins. Our results suggest that when a bacterium presents membrane proteins that make it compete better in its environment, and these proteins are in turn receptors for specific phages, it would be forced to acquire CRISPR-Cas immunity systems to avoid infection by these phages.

bioinformatics↗