NPJ Systems Biology and Applications最新文献

While reductionism has advanced biology and medicine, it fosters a fragmented understanding of health, ill-suited to modern challenges like chronic and systemic diseases. Systems biology offers a new perspective, framing biological entities within interconnected networks. Using French medical education as an example, we argue that systems thinking should be foundational, not optional. Integrating systems biology early in medical education can better prepare future physicians for biomedical complexity and precision medicine.

Antibiotic persistence, characterized by a dormant subpopulation of bacterial cells that causes chronic and recurrent infections, remains poorly understood despite being recognized nearly a century ago. Toxin-antitoxin (TA) systems, which include a toxin and an antitoxin, are promising candidates for elucidating persister formation. We present the first theoretical model of persister formation driven by type I TA systems, in which the antitoxin is a small RNA molecule. Our analyses and simulations reveal two steady states-low toxin (normal growth) and high toxin (persistence)-with stochastic switching between them. Bistability requires both positive and negative feedback mediated by inhibition of antitoxin degradation. We derive stability diagrams that map mechanistic properties to system dynamics. The model suggests that while type I TA systems may not produce persisters under normal conditions, they can enter a bistable regime under stress, such as antibiotic exposure or nutrient limitation, leading to increased toxin expression or slower growth. Moreover, transiently slow-growing cells can be stabilized as long-living persisters through bistable TA dynamics. Using a cusp catastrophe surface, we identify distinct roles for two toxin inhibition mechanisms in modulating steady states and hysteresis. These findings provide a mechanistic basis for experimental observations and a framework for future studies.

Cell fusion generates hybrid cells with unique traits. To understand the transcriptional and signaling alterations after fusion, we analyzed a published single-cell RNA-sequencing dataset of fused murine cardiomyocytes (mHL1) and mesenchymal stromal/stem cells (mMSC). Our analysis showed that fused cells exhibit a transcriptional trajectory suggesting a rapid change that stabilizes over time. We observed asymmetric plasticity. Initially, at Day 1, fusion hybrids resembled mMSCs (mesenchymal reprogramming), but by Day 3, their gene expression shifted to resemble mHL1 cells (myogenic reprogramming). Our analysis also identified distinct transcriptional subpopulations, including a subset enriched for tenascin (extracellular matrix remodeling), accompanied by dynamic changes in cell adhesion and intercellular communication. We also saw a significant shift in signaling pathways over time. At Day 1, Wnt and Melanogenesis (regenerative/antioxidant) signaling were downregulated. By Day 3, stress resistance and cellular adaptation pathways became enriched. Gene regulatory network analysis revealed key changes in master regulators; genes associated with chromatin remodeling (Hmga2), circadian rhythm (Arntl), and mesenchymal identity (Prrx1) became more active by Day 3. Collectively, our findings demonstrate that cell fusion is a dynamic reprogramming process, where evolving gene regulatory and signaling networks generate novel hybrid cell states, creating cellular diversity.

Epileptic seizures result from abnormal synchronous neuronal firing caused by an imbalance between excitatory and inhibitory neurotransmission. While most seizures are self-limiting, those lasting over five minutes, termed status epilepticus, require medical intervention. Benzodiazepines, the first-line treatment, terminate seizures by enhancing GABAergic inhibition, but fail in approximately 36% of cases. In this paper, we employ a neural mass framework to investigate how different interventions influence brain dynamics and facilitate seizure termination. As seizures are characterized by persistent firing, we extend the classic Wilson-Cowan framework by introducing a term called sustenance which encodes factors that promote or discourage perpetual firing. The resulting model captures transitions between normal activity and seizure and provides a tractable framework for analysing diverse pathophysiological mechanisms. We first show how various dysfunctions-such as hyperexcitation, depletion of inhibitory neurotransmitters, and depolarizing GABAergic transmission-can all give rise to seizures, with overlapping but distinct dynamics. Building on this foundation, we turn to the central question of intervention: how different treatments act on these mechanisms to terminate seizures. We find that while enhancing GABAergic inhibition is generally effective, it fails when GABA becomes depolarizing. In such cases, interventions like levetiracetam that suppress sustained excitatory activity remain effective. These findings highlight the importance of aligning interventions to the specific underlying dysfunction for effective seizure termination.

Spatial omics (SO) produces high-definition mapping of subcellular molecules within tissue samples. Mapping transcripts to anatomical regions requires segmentation, but this remains challenging in tissue cross-sections with tubular structures like axons in peripheral nerve or spinal cord. Neural networks could address misidentification but are hindered by the need for extensive human annotations. We present SiDoLa-NS (Simulate, Don't Label-Nervous System), an image-driven (top-down) approach to SO analysis in the nervous system. We utilize biophysical properties of tissue architectures to design synthetic images of tissue samples, eliminating reliance on manual annotation and enabling scalable training data generation. With synthetic samples, we trained supervised instance segmentation convolutional neural networks (CNNs) for nucleus segmentation, achieving precision and F1-scores>0.95. We further identify macroscopic tissue structures in mouse brain (mAP₅₀=0.869), spinal cord (mAP₅₀=0.96), and pig sciatic nerve (mAP₅₀=0.995). This framework sets the stage for transferable models across species and tissue architectures-accelerating SO applications in neuroscience and beyond.

Glycyrrhiza uralensis, a key component of over 70% of traditional herbal medicines (Kampo) in Japan, exhibits diverse pharmacological effects, including immunoregulation, anti-tumor, and antioxidant properties. Despite over 300 identified compounds, the molecular mechanisms remain unclear due to the chemical diversity. Here, we performed a multiomics analysis incorporating untargeted hydrophilic metabolomics, lipidomics, and phosphoproteomics to elucidate the mechanisms distinguishing the G. uralensis extract (GU) from a single bioactive compound, isoliquiritigenin (ILG). Time-course analyses of lipopolysaccharide (LPS)-stimulated RAW264.7 cells under four conditions (control, LPS(+), LPS(+)/ILG(+), and LPS(+)/GU(+)) quantified 182 hydrophilic metabolites, 381 lipids, and 13,211 phosphopeptides. Both ILG(+) and GU(+) attenuated inflammatory signatures characterized by elevated glycolytic intermediates, succinate, citrulline, triacylglycerols, and cholesteryl esters. A multiset partial least squares technique identified sirtuin (SIRT) 1/2 phosphorylation and altered nicotinamide adenine dinucleotide metabolism specific to ILG(+). SIRT2 inhibition abolished ILG's suppression of interleukin-6 (IL-6). Furthermore, GU(+) uniquely increased γ-aminobutyric acid (GABA) and 4-guanidinobutyric acid via endogenous synthesis by glutamic acid decarboxylase. Exogenous GABA reduced IL-6 and IL-1β expression, and its co-administration with ILG enhanced anti-inflammatory effects. This study demonstrates that multiomics can elucidate the synergistic anti-inflammatory actions of G. uralensis, highlighting endogenous GABA production as a key contributor to ILG-mediated immunomodulation.

Genome-scale metabolic models (GEMs) are valuable tools for investigating healthy and disease states, but often lack the specificity to capture context-dependent metabolic adaptations. Tailoring GEMs using transcriptomic data is crucial for studying these context-specific variations by accurately identifying active metabolic reactions. This study introduces an algorithm called 'Localgini', which uses the Gini coefficient to quantify gene expression variability across samples, enabling precise identification of active reactions for context-specific models (CSMs). To evaluate Localgini, CSMs were generated using six different model extraction methods (MeMs) for NCI-60 cancer cell lines and human tissue datasets. Localgini-based CSMs better represent housekeeping functionalities and known metabolic pathways. Moreover, Localgini-generated active reaction sets require minimal support from the MeMs to build the CSMs. Localgini minimizes variability across CSMs built with different MeMs and the same gene expression data. Overall, by incorporating gene expression heterogeneity, Localgini provides an accurate method for constructing CSMs.

Although two-thirds of cancers arise from loss-of-function mutations in tumor suppressor genes, there are few approved targeted therapies linked to these alterations. Synthetic lethality offers a promising strategy to treat such cancers by targeting vulnerabilities unique to cancer cells with these mutations. To identify clinically relevant synthetic lethal interactions, we analyzed genome-wide CRISPR/Cas9 knock-out (KO) viability screens from the Cancer Dependency Map and evaluated their clinical relevance in patient tumors through mutual exclusivity, a pattern indicative of synthetic lethality. Indeed, we found significant enrichment of mutual exclusivity for interactions involving cancer driver genes compared to non-driver mutations. To identify therapeutic opportunities, we integrated drug sensitivity data to identify inhibitors that mimic the effects of CRISPR-mediated KO. This approach revealed potential drug repurposing opportunities, including BRD2 inhibitors for bladder cancers with ARID1A mutations and SIN3A-mutated cell lines showing sensitivity to nicotinamide phosphoribosyltransferase (NAMPT) inhibitors. However, we discovered that pharmacological inhibitors often fail to phenocopy KO of matched drug targets, with only a small fraction of drugs inducing similar effects. This discrepancy reveals fundamental differences between pharmacological and genetic perturbations, emphasizing the need for approaches that directly assess the interplay of loss-of-function mutations and drug activity in cancer models.

Predicting how gut microbial communities assemble and change requires models that capture the underlying mechanisms driving interspecies interactions, not just taxonomic correlations. We present SIMBA, a simulation-augmented graph neural network that integrates mechanistic insights from metabolic simulations with edge-aware graph transformers to predict microbial community composition. Using a high-fiber dietary cohort mapped to metabolic networks, we ran thousands of pairwise simulations to infer cross-feeding probabilities, pathway activity fingerprints, and microbe-microbe functional similarity. These signals instantiate a global microbe-metabolite-pathway graph for learning. A custom heterogeneous graph transformer incorporates scalar edge attributes into attention. It is trained through a multi-stage pipeline combining self-supervised learning, supervised pretraining on simulated graphs, and fine-tuning on experimental microbial abundance data. Each individual's microbiome is represented as a sample-specific instantiation of the shared mechanistic graph derived from metabolic simulations, where only the set of microbes detected in that individual varies. SIMBA learns from this mechanistic prior to predict microbial presence and relative abundance across individuals, enabling hypothesis-driven exploration of microbial ecosystems.