Reconstructing gene regulatory networks (GRNs) from single-cell RNA sequencing (scRNA-seq) data is fundamental for understanding cellular dynamics at the molecular level, but requires sophisticated workflows. Here, we introduce CellCraft, a web-based application designed to streamline GRN inference. CellCraft integrates multiple GRN reconstruction tools, including TENET, within a unified web application featuring an intuitive graphical user interface (GUI). Notably, CellCraft provides a visual programming interface that simplifies the design and execution of complex multi-step analyses, thereby enhancing accessibility and facilitating the visualization and interpretation of computational experiments. Furthermore, its modular plugin architecture ensures extensibility, enabling the incorporation of newly developed single-cell analysis algorithms. Consequently, CellCraft provides a user-friendly and extensible application for integrative GRN analysis of scRNA-seq datasets. CellCraft is available on GitHub at https://github.com/cxinsys/cellcraft. The source code has been archived on Zenodo at 10.5281/zenodo.17865848. Supplementary data is available at Bioinformatics online.

Single-cell expression quantitative trait loci (sceQTL) mapping offers a powerful approach for understanding gene regulation and its heterogeneity across cell types and states. It has profound applications in genetics and genomics, particularly causal gene regulatory network (cGRN) inference to unravel the molecular circuits governing cell identity and function. However, computational scalability remains a critical bottleneck for sceQTL mapping, prohibiting thorough benchmarking and optimization of statistical accuracy. We present airqtl, an efficient method to overcome these challenges through algorithmic advances and efficient implementations of linear mixed models. Airqtl achieves superior time complexity and over 108 times of acceleration, enabling objective method benchmarking and optimization. Airqtl offers de novo inference of robust, experimentally validated cell state-specific cGRNs that reflect perturbation outcomes. Our results dissect the drivers of cGRN heterogeneity and underscore the value of natural genetic variations in primary human cell types for biologically relevant single-cell cGRN inference.

Brassica napus L. (rapeseed) represents one of the world's most important oilseed crops, yet its yield is increasingly impacted by heat-a threat that is intensifying under climate change. Here, we employed an integrative framework combining RNA‑Seq meta‑analysis, weighted gene co‑expression network analysis (WGCNA), and gene regulatory network (GRN) inference to dissect the transcriptional programs activated during the response of rapeseed to heat stress. WGCNA partitioned the transcriptome into three principal modules-green, black, and brown-with genes in the black and brown modules exhibiting predominantly decreased expression under heat, whereas the green module was largely characterized by upregulation. The green module promoted energy conservation, transcriptome flexibility through alternative splicing, endoplasmic reticulum remodeling, proteostasis via chaperone activity, redox balance, and reduced ABA sensitivity to enhance transpirational cooling. In contrast, the black and brown modules, largely downregulated, restricted carbon fixation, cyclic electron flow, chlorophyll turnover, and PSII repair while sustaining stomatal opening, highlighting a trade-off between photosynthetic efficiency and thermotolerance. Within GRNs, upregulated hub TFs reinforced chaperone function, ROS detoxification, (post-)transcriptional reprogramming, and photosynthetic support whereas downregulated hub TFs suppressed photosynthesis, metabolism, growth, transport, and basal immunity thereby redirecting resources toward essential heat stress responses. Our results illuminate the key regulatory mechanisms and co‑expressed gene clusters that orchestrate adaptive response of rapeseed to heat, offering valuable molecular targets for the development of varieties with improved tolerance to heat stress conditions.

Gene regulatory network (GRN) provides critical insights into the molecular mechanisms that govern cellular processes and disease pathogenesis, facilitating the identification of key regulatory factors and the discovery of potential therapeutic targets. Although numerous methods have been proposed to infer GRN from single-cell RNA sequencing (scRNA-seq) data, GRN inference remains challenging due to the inherent sparsity of scRNA-seq data and the naturally sparse connectivity of GRN. To address these challenges, this study proposes the Multimodal Adaptive GRN Inference Constructor (MAGIC), a method that improves GRN inference by aligning and integrating gene expression data, sequence information, and semantic features. Specifically, gene expression features reflect gene activity within cells, gene sequence features offer structural insights at the DNA level, and gene semantic features encapsulate functional meaning by leveraging biological knowledge bases. Furthermore, a consensus similarity network is constructed from multimodal gene similarity networks and integrated with known GRN to form a dual-topology network. To address the issue of sparse connectivity in GRN, a shared graph attention weight alignment module is employed. Following this, a Knowledge-Aware Multimodal Fusion Module is introduced to effectively integrate multimodal features by leveraging prior knowledge, thereby alleviating the inherent sparsity of scRNA-seq data. Finally, the fused features are used to infer GRNs. MAGIC achieved an average AUROC of 0.839 across seven scRNA-seq data sets using four types of ground-truth networks, outperforming other state-of-the-art models. Further analysis of two spatial transcriptomic data sets, bladder and breast cancer, demonstrates the robustness of MAGIC and its ability to uncover potential associations between transcription factors (TFs) and their target genes. MAGIC is publicly available at https://github.com/ydkvictory/MAGIC.

Single-cell transcriptomics reveals the functional trajectories of the endocarp cell clusters in the postharvest fruit senescence process of Hylocereus undatus.

Gene regulatory network (GRN) inference is fundamental to understanding the regulatory architecture underlying cellular processes. Accurate reconstruction of cell-type-specific GRNs is therefore essential for elucidating the mechanisms that govern cellular identity, development, and disease. However, inferring GRNs from single-cell RNA sequencing data remains challenging due to data sparsity, noise, and the intrinsic complexity of gene regulation. Here, RegGAIN is presented, a novel deep learning-based model designed to infer GRNs from single-cell transcriptomic data. RegGAIN employs self-supervised contrastive learning to maximize consistency of gene embeddings across perturbed graph views. To characterize regulatory directionality and capture the distinct regulator- and target-driven patterns simultaneously, it leverages separate encoders to learn dual-role representations for each gene. Comprehensive evaluations demonstrate that RegGAIN achieves accurate and robust GRN reconstruction, consistently outperforming existing methods. The biological relevance of the predicted regulatory interactions is further validated using external epigenetic data. Moreover, RegGAIN enables the discovery of GRN rewiring, revealing condition-specific and temporally dynamic regulatory programs. Together, RegGAIN offers a powerful and generalizable framework for GRN inference, paving the way for deeper insights into cellular regulation across diverse biological contexts.

ABSTRACTSpinal cord aging is a critical physiological process that compromises central nervous system (CNS) homeostasis and plasticity. Exercise, as a systemic intervention with broad health benefits, has been shown to delay neurodegeneration and preserve tissue function; however, its impact on dynamic cellular lineage evolution and intercellular communication within the aging spinal cord remains poorly characterized. In this study, we employed single‐nucleus RNA sequencing (snRNA‐seq) to construct a high‐resolution cellular atlas of the mouse spinal cord under young, aged, and aerobic exercise‐intervened conditions. By integrating unsupervised clustering, cell proportion analysis, pseudotime trajectory reconstruction, gene regulatory network (GRN) inference, and intercellular communication mapping, we systematically characterized transcriptional and cellular alterations associated with aging and their modulation through exercise. Aging induced pronounced shifts in cell‐type composition and subpopulation structures, which were partially reversed by exercise intervention. Pseudotime analyses of oligodendrocytes, astrocytes, and microglia revealed that exercise remodeled their differentiation trajectories and restored functional states associated with myelin formation, metabolic homeostasis, and inflammation control. GRN analysis identified several key regulators whose centrality and expression were disrupted during aging but reestablished after exercise, suggesting a recovery of transcriptional network organization. Furthermore, intercellular communication analysis revealed reduced signaling strength and connectivity during aging, particularly within gap junction pathways, which were partially restored by exercise, indicating improved cellular coordination. Together, these findings provide a comprehensive single‐cell landscape of the aging spinal cord and demonstrate that exercise reprograms cellular lineages and regulatory networks, offering mechanistic insights into how it mitigates CNS aging and preserves neural function.

During neurogenesis, signaling molecules and transcription factors (TFs) pattern neural progenitors across space and time to generate the numerous cell types that constitute neural circuits. In postmitotic neurons, these identities are established and maintained by a class of TFs known as terminal selectors (tsTFs). However, it remains largely unclear how the tsTF combinations are specified, and how they then coordinate the type-specific differentiation programs of each neuron. To investigate these regulatory mechanisms, we performed simultaneous single-cell RNA and ATAC sequencing on the Drosophila optic lobes at four stages of their development and identified over 250 distinct cell types. We characterized the common cis-regulatory features of neuronal enhancers and performed comprehensive inference of gene regulatory networks across cell types and stages. Our results reveal cell-type and stage-specific enhancers of many neuronal genes and the cooperative actions of pan-neuronal and tsTFs on these enhancers. We show that the same effector genes are often regulated by different tsTF combinations in different neurons. During neurogenesis, most temporal TFs and the spatial TF Vsx1 are regulated by different enhancers before and after neurons become postmitotic, allowing them to be re-utilized as tsTFs independently from their roles in progenitors. As proof of concept, we genetically dissected the regulation of Vsx1 as a tsTF by different spatial and temporal patterning mechanisms through lineage-specific enhancers.

Inferring gene regulatory networks (GRNs) is essential for understanding biological regulation. Although numerous deep learning approaches have been developed for GRN inference, most require large amounts of labeled data. We present Meta-TGLink, a structure-enhanced graph meta-learning model for few-shot GRN inference. By formulating GRN inference as a link prediction task, Meta-TGLink captures transferable regulatory patterns while reducing dependence on extensive labeled datasets. The model combines graph neural networks with Transformer architectures to integrate relational and positional information, thereby improving predictive performance under data-scarce conditions. Experiments on real datasets demonstrate its superiority over state-of-the-art baselines, particularly in cross-domain few-shot scenarios.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13059-025-03860-8.

Abstract INTRODUCTION Malignant gliomas follow two distinct natural histories: de novo high grade tumors such as glioblastoma, or lower grade tumors with a propensity to transform into high grade disease. Despite differences in tumor genotype, both entities converge on a common histologically aggressive phenotype, and the basis for this progression is unknown. Glioma associated macrophages (GAM) have been implicated in this process, however GAMs are ontologically and transcriptionally diverse, rendering the identification and isolation of pathogenic subpopulations challenging. OBJECTIVE Since macrophage contextual gene programs are orchestrated by transcription factors acting on cis-acting promoters and enhancers in gene regulatory networks (GRN), we hypothesized that functional subpopulations of GAMs can be resolved through GRN inference. Method IDH mutant/wildtype gliomas and control normal brain samples were collected for single cell RNA +ATAC sequencing to derive GRNs. Transcription factor network analysis identified cell surface markers to isolate subpopulations for functional assays, with tissue, in vitro and in vivo correlates. RESULTS A subpopulation of human GAMs can be defined by a GRN centered around the Activator Protein-1 transcription factor FOSL2 preferentially enriched in high grade tumors. Using this GRN we nominate ANXA1 and HMOX1 as surrogate cell surface markers for activation, thus permitting prospective isolation and functional validation in human GAMs. These cells, termed malignancy associated GAMs (mGAMs) are pro-invasive, pro-angiogenic, pro-proliferative, possess intact antigen presentation but skew T-cells towards a CD4+FOXP3+ phenotype under hypoxia. Ontologically, mGAMs share somatic mitochondrial mutations with peripheral blood monocytes, and their presence correlates with high grade disease irrespective of underlying tumor mutation status. Furthermore, spatio-temporally mGAMs occupy distinct metabolic niches; mGAMs directly induce proliferation and mesenchymal transition of low grade glioma cells and accelerate tumor growth in vivo upon co-culture. Finally mGAMs are preferentially enriched in patients with newly transformed regions in human gliomas. CONCLUSION mGAMs play a pivotal role in glioma progression and represent a potential therapeutic target in human high-grade glioma.

This study aimed to identify prognostic features in high-grade serous ovarian cancer (HGSOC) through the application of gene regulatory network (GRN) inference with single-cell RNA-sequencing (scRNA-seq) profiles. To achieve this goal, we developed a workflow comprising scRNA-seq analysis, metacell construction, GRN inference, and a binary classification task for prognosis prediction. We curated 118,173 cells from HGSOC patients in three conditions (Before-chemotherapy, After-chemotherapy, and control samples) from previous studies, and then constructed 1,211 metacells. GRN inference analysis revealed 312 regulons, each consisting of one transcription factor and its targeted features. For prognosis evaluation, we used bulk RNA-seq data covering 342 HGSOC patients from The Cancer Genome Atlas (TCGA) and defined a binary outcome of overall survival ≥ 2 years from initial diagnosis, with censored cases at last follow-up assigned to the appropriate class by observed time. We prioritized the features of the TCGA data based on regulon information and differentially expressed features extracted from the metacell data. Our results demonstrated that regulon-based prognostic features were more effective than differential expression-based features in both Before-chemotherapy and After-chemotherapy groups. Our framework can be generalized to other types of cancer when single-cell data for GRN inference and bulk RNA-seq data with clinical outcomes are available.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-22937-9.

Oscillatory processes such as the cell cycle play critical roles in cell fate determination and disease development. Yet, most current gene regulatory network (GRN) inference methods are based on gene-gene correlations or temporal progression, not adequately accounting for the recurrence in cyclic processes. We hypothesize that constraining the continuous ordering of relative positions along the cell cycle can enhance GRN inference accuracy of cell cycle regulation. To test performance, we evaluated eight representative methods and applied three of them to a mouse retinal progenitor single-cell gene expression dataset [1]. Incorporating cell cycle positions inferred by Tricycle [2] led to significant improvements compared against using experimental times, particularly for early progenitor cells that been hypothesized to be more intrinsically driven by cell cycle regulation. These findings highlight the promise of integrating oscillatory processes into causal inference frameworks to advance our understanding of gene regulation.

Dissecting pirtobrutinib resistance in Mantle Cell Lymphoma through single-cell multi-omics

Determining mechanistic models of gene regulation, especially underlying phenotypic variation, is a central goal of both mathematical biology and modern evolutionary biology. However, several challenges, involving both common characteristics of experimental data and the model development process, remain that limit the discovery of general principles. Even the highest-quality experimental data come with challenges. There are always sources of noise, a limit to how often we can measure the system in time, and it is impossible to measure all the relevant states that participate in the full underlying complexity. Additionally, there are usually sources of uncertainty in the underlying biological mechanisms, which give rise to multiple competing model structures. We walk through a case study involving inference of a regulatory network structure involved in a developmental decision in the nematode, \textit{Pristonchus pacificus}. In this study, we fit 13,824 distinct regulatory network models to gene expression data from three experimental conditions to determine which regulatory features are supported by the data. We discover \textit{model sets}, or collections of models with shared regulatory network features that best fit the data, for each of the three experiments we considered, and identify a regulatory network in the intersection of the three model sets. This model describes the data across the experimental conditions and exhibits a high degree of positive regulation and interconnectivity between the key regulators, \textit{eud-1}, \textit{sult-1}, and \textit{nhr-40}. While the biological results are specific to the molecular biology of development in \textit{Pristonchus pacificus}, the comparative modeling framework introduced here can be applied to other systems of gene regulation in an evolutionary developmental context.

Abstract The inference of gene regulatory networks (GRNs) is critical for understanding the regulatory mechanisms underlying cellular development, functional specialization, and disease progression. Predicting regulatory gene interactions—often framed as a link prediction task—is a foundational step toward modeling cellular behavior. However, GRN inference from gene coexpression data alone is limited by noise, low interpretability, and difficulty in capturing indirect regulatory signals. Additionally, challenges such as data sparsity, nonlinearity, and complex gene interactions hinder accurate network reconstruction. To address these issues, we propose, a novel graph transformer (GT) based framework (GT-GRN) that enhances GRN inference by integrating multimodal gene embeddings. Our method combines three complementary sources of information: (i) autoencoder-based embeddings, which capture high-dimensional gene expression patterns while preserving biological signals; (ii) structural embeddings, derived from previously inferred GRNs and encoded via random walks and a Bidirectional Encoder Representations from Transformers (BERT) based language model to learn global gene representations; (iii) positional encodings, capturing each gene’s role within the network topology . These heterogeneous features are fused and processed using a GT, allowing the joint modeling of both local and global regulatory structures. Experimental results on benchmark datasets show that GT-GRN outperforms existing GRN inference methods in predictive accuracy and robustness. Furthermore, it reconstructs cell-type-specific GRNs with high fidelity and produces gene embeddings that generalize to other tasks such as cell-type annotation.

An approach to inferring gene regulatory networks via boolean modeling and feature selection.

Advancements in spatial omics permit spatially resolved measurements across several biological modalities. The high cost of acquiring co-profiled multimodal data limits the analysis. This underscores the necessity for computational methods to integrate unpaired spatial multi-omics data and perform cross-modal predictions on single-modality data. The integration of spatial omics is challenging due to typically low signal-to-noise ratios. Here we introduce SWITCH (Spatially Weighted Multi-omics Integration and Cross-modal Translation with Cycle-mapping Harmonization), a deep generative model for spatial multi-omics integration. SWITCH presents a cycle-mapping mechanism that produces dependable cross-modal translations without requiring additional paired data. These cross-modal translations function as pseudo-pairs to provide supplementary signals. Systematic evaluations demonstrate that SWITCH outperforms existing methods in terms of integration accuracy and achieves more precise spatial domain delineation, resolving brain cortical structures at higher resolution. The reliability of cross-modal translations was validated, facilitating various downstream analyses such as differential analysis, trajectory inference and gene regulatory network inference.

Knowing gene regulatory networks (GRNs) is important for understanding various biological mechanisms. In this paper, we present a method, QWENDY, that uses single-cell gene expression data measured at four time points to infer GRNs. Based on a linear gene expression model, it solves the transformation of the covariance matrices. Unlike its predecessor WENDY, QWENDY avoids solving a non-convex optimization problem and produces a unique solution. We test the performance of QWENDY on three experimental data sets and two synthetic data sets. Compared to previously tested methods on the same data sets, QWENDY ranks the first on experimental data, although it does not perform well on synthetic data.

The interplay between transcription factors (TFs) and regulatory elements (REs) drives gene transcription, forming gene regulatory networks (GRNs). Advances in single-cell technologies now enable simultaneous measurement of RNA expression and chromatin accessibility, offering unprecedented opportunities for GRN inference at single-cell resolution. However, heterogeneity across omics layers complicates regulatory feature extraction. We present scTFBridge, a multi-omics deep generative model for GRN inference. scTFBridge disentangles latent spaces into shared and specific components across omics layers. By integrating TF-motif binding knowledge, scTFBridge aligns shared embeddings with specific TF regulatory activities, enhancing biological interpretability. Using explainability methods, scTFBridge computes regulatory scores for REs and TFs, enabling robust GRN inference. Our results further demonstrate that scTFBridge can identify cell-type-specific susceptibility genes and distinct regulatory programs, providing insights into gene regulation mechanisms at the single-cell level.

Gene regulatory network (GRN) inference is a central task in systems biology. However, due to the noisy nature of gene expression data and the diversity of regulatory structures, accurate GRN inference remains challenging. We hypothesize that integrating multi-source features and leveraging an attention mechanism that explicitly captures graph structure can enhance GRN inference performance. Based on this, we propose GTAT-GRN, a deep graph neural network model with a graph topological attention mechanism that fuses multi-source features. GTAT-GRN includes a feature fusion module to jointly model temporal expression patterns, baseline expression levels, and structural topological attributes, improving node representation. In addition, we introduce the Graph Topology-Aware Attention Network (GTAT), which combines graph structure information with multi-head attention to capture potential gene regulatory dependencies. We conducted comprehensive evaluations of GTAT-GRN on multiple benchmark datasets and compared it with several state-of-the-art inference methods, including GENIE3 and GreyNet. The experimental results show that GTAT-GRN consistently achieves higher inference accuracy and improved robustness across datasets. These findings indicate that integrating graph topological attention with multi-source feature fusion can effectively enhance GRN reconstruction.