Frontiers in systems biology最新文献

Prioritizing actionable drug targets is a critical challenge in cancer research, where high-dimensional genomic data and the complexity of tumor biology often hinder effective prioritization. To address this, we developed GETgene-AI, a novel computational framework that integrates network-based prioritization, machine learning, and automated literature analysis to prioritize and rank potential therapeutic targets. Central to GETgene-AI is the G.E.T. strategy, which combines three data streams: mutational frequency (G List), differential expression (E List), and known drug targets (T List). These components are iteratively refined and ranked using the Biological Entity Expansion and Ranking Engine (BEERE), leveraging protein-protein interaction networks, functional annotations, and experimental evidence. Additionally, GETgene-AI incorporates GPT-4o, an advanced large language model, to automate literature-based ranking, reducing manual curation and increasing efficiency. In this study, we applied GETgene-AI to pancreatic cancer as a case study. The framework successfully prioritized high-priority targets such as PIK3CA and PRKCA, validated through experimental evidence and clinical relevance. Benchmarking against GEO2R and STRING demonstrated GETgene-AI's superior performance, achieving higher precision, recall, and efficiency in prioritizing actionable targets. Moreover, the framework mitigated false positives by deprioritizing genes lacking functional or clinical significance. While demonstrated on pancreatic cancer, the modular design of GETgene-AI enables scalability across diverse cancers and diseases. By integrating multi-omics datasets with advanced computational and AI-driven approaches, GETgene-AI provides a versatile and robust platform for accelerating cancer drug discovery. This framework bridges computational innovations with translational research to improve patient outcomes.

In silico clinical trials that simulate individualized mechanisms of action offer a powerful approach to assess drug efficacy across large and diverse patient populations, while also enabling the identification of predictive biomarkers. In this study, we conducted an in silico clinical trial of first-line, single-agent regorafenib in 399 elderly patients with metastatic colorectal cancer (mCRC). Individualized network-based models were constructed using patient-specific differential transcriptomic profiles and employed to simulate the target-specific effects of regorafenib. From this analysis, we identified both predictive and mechanistic biomarkers of treatment response. Notably, four proteins-MARK3, RBCK1, LHCGR, and HSF1-emerged as dual biomarkers, showing associations with both response mechanisms and predictive potential. Three of these (MARK3, RBCK1, and HSF1) were validated in an independent cohort of mCRC patients and were also found to be targets of previously reported regorafenib-predictive miRNAs. This study demonstrates a novel systems biology strategy for evaluating drug response in silico, leveraging transcriptomic data to simulate individual treatment outcomes and uncover clinically relevant biomarkers. Our findings suggest that such approaches may serve as valuable complements to traditional clinical trials for assessing drug efficacy and guiding precision oncology.

Collaborations between industry leaders and academia are crucial for advancing systems biology education and training. This article explores opportunities for partnerships to enhance the educational landscape and develop a workforce skilled in systems modelling, particularly for quantitative systems pharmacology (QSP) in drug development. Companies with a strong focus on innovation frequently explore collaborative ventures involving joint research, co-designed curricula, and specialized training programs. These partnerships provide students and researchers with insights into real-world applications of systems biology and QSP. We explicitly review success criteria for collaboration at MSc and PhD levels, discuss earlier pipeline considerations, and carefully balance the roles, incentives, and challenges for both academia and industry in collaborative ventures. Challenges in aligning academic and industry objectives exist, including resource allocation and intellectual property concerns. However, the importance of training skilled systems biologists for advancing drug discovery and development outweighs these challenges. The article concludes by highlighting successful industry-academia partnerships and offering recommendations for optimizing collaborations to meet the evolving needs of systems biology education and drive innovation in pharmaceutical research.

The microbiome represents a complex community of trillions of microorganisms residing in various body parts and plays critical roles in maintaining host health and wellbeing. Understanding the interactions between microbiota and their host offers valuable insights into potential strategies for promoting health, including microbiome-targeted interventions. We have created MicrobiomeKG, a knowledge graph for microbiome research, that bridges various taxa and microbial pathways with host health. This novel knowledge graph derives algorithmically generated knowledge assertions from the supplementary tables that support published microbiome papers. By identifying knowledge assertions from supplementary tables and expressing them as knowledge graphs, we are casting this valuable content into a format that is ideal for hypothesis generation. To address the high heterogeneity of study contexts, methodologies, and reporting standards, we leveraged neural networks to implement a standardized edge scoring system, which we use to perform centrality analyses. We present three example use cases: linking helminth infections with non-alcoholic fatty-liver disease via microbial taxa, exploring connections between the Alistipes genus and inflammation, and identifying the Bifidobacterium genus as the most central connection with attention deficit hyperactivity disorder. MicrobiomeKG is deployed for integrative analysis and hypothesis generation, both programmatically and via the Biomedical Data Translator ecosystem. By bridging data gaps and facilitating the discovery of new biological relationships, MicrobiomeKG will help advance personalized medicine through a deeper understanding of the microbial contributions to human health and disease mechanisms.

A representation of the cause-effect mechanism is needed to enable artificial intelligence to represent how the world works. Bayesian Networks (BNs) have proven to be an effective and versatile tool for this task. BNs require constructing a structure of dependencies among variables and learning the parameters that govern these relationships. These tasks, referred to as structural learning and parameter learning, are actively investigated by the research community, with several algorithms proposed and no single method having established itself as standard. A wide range of software, tools, and packages have been developed for BNs analysis and made available to academic researchers and industry practitioners. As a consequence of having no one-size-fits-all solution, moving the first practical steps and getting oriented into this field is proving to be challenging to outsiders and beginners. In this paper, we review the most relevant tools and software for BNs structural and parameter learning to date, with a focus on causal discovery tools, providing our subjective recommendations directed to an audience of beginners. In addition, we provide an extensive easy-to-consult overview table summarizing all software packages and their main features. By improving the reader's understanding of which available software might best suit their needs, we improve accessibility to the field and make it easier for beginners to take their first step into it.

Mathematical models for cellular systems have become more and more important for understanding the complex interplay between metabolism, signalling, and gene expression.In this manuscript, starting from the well-known flux balance analysis, tools and methods are summarised and illustrated by various examples that describe the way to models with kinetics for individual reactions steps that are finally self-contained. While flux analysis requires known (measured) input fluxes, self-contained (or self-sustained) models only get information on concentrations of environmental components. Kinetic reaction laws, feedback structures, and protein allocation then determine the temporal output of all intracellular metabolites and macromolecules. Emphasis is placed on (i) mass conservation, a crucial system property frequently overlooked in models incorporating cellular structures like macromolecular structures like proteins, RNA, and DNA, and (ii) thermodynamic constraints which further limit the solution space. Matlab Live Scripts are provided for all simulation studies shown and additional reading material is given in the appendix.

miRNAs are promising diagnostic biomarkers. These small RNA molecules are always present in the human body but become dysregulated when a person develops certain diseases. Although the detection of these biomarkers in cell-free tests is ongoing work, current systems often focus solely on detecting the presence or absence of a specific miRNA, rather than the miRNAs concentration. Thus, these tests may miss relative changes in miRNA concentration when disease-induced dysregulation occurs. This work, part of the WUR iGEM 2024 project (miRADAR), aimed to address this gap by incorporating an miRNA concentration-dependent threshold mechanism in a cell-free diagnostic test. In this system, continuous miRNA input concentrations need to be converted into a binary output signal, classifying the miRNA concentration as healthy (no output signal) or indicative of disease (strong output signal). To aid the experimental engineering of the test, here we use mathematical models to evaluate and assess different candidate networks. We apply a previously published multi-objective optimisation strategy to obtain designs that satisfy relevant constraints, such as low basal expression, high readout levels, and steep switching behaviour between low and high input miRNA concentrations. Models for three different biological mechanisms were compared based on their ability to generate the desired binary output signal. One approach used three-node protein networks (such as feed-forward loops), while the other two utilised RNA-based toehold systems. Overall, the toehold-mediated strand displacement systems demonstrated the most potential for experimental implementation. These systems are believed to be less burdensome in a cell-free environment, can be more readily engineered for new miRNA sequences, and showed high detection accuracy. Based on our results, we discuss how the inclusion of sequence-specific parameters could expand the design space of our mathematical models and how careful engineering of optimisation criteria is required to evaluate designs. Ultimately, our model-based study highlights that toehold-mediated strand displacement networks have the potential to be efficient miRNA detection systems for biosensing tools in the future.

Irrational antibiotic use contributes to the development of antibiotic resistance in bacteria, which is a major cause of healthcare-associated infections globally. Molecular research has shown that multiple resistance frequently develops from the uptake of pre-existing resistance genes, which are subsequently intensified under selective pressures. Resistant genes spread and are acquired through mobile genetic elements which are essential for facilitating horizontal gene transfer. MGEs have been identified as carriers of genetic material and are a significant player in evolutionary processes. These include insertion sequences, transposons, integrative and conjugative elements, plasmids, and genomic islands, all of which can transfer between and within DNA molecules. With an emphasis on pathogenic bacteria, this review highlights the salient features of the MGEs that contribute to the development and spread of antibiotic resistance. MGEs carry non-essential genes, including AMR and virulence genes, which can enhance the adaptability and fitness of their bacterial hosts. These elements employ evolutionary strategies to facilitate their replication and dissemination, thus enabling survival without positive selection for the harboring of beneficial genes.

By 2050, global population growth will significantly increase food demand, placing additional pressure on agriculture, a sector already vulnerable to climate change. Traditional approaches like fertilizers and pesticides have helped boost yields but are increasingly seen as unsustainable. As bioengineering becomes more accessible, engineered soil microorganisms are emerging as promising alternatives. However, their application in the rhizosphere is often limited by poor survivability and the high metabolic cost of expressing heterologous genes without appropriate regulation. To address this, we developed a microbial whole-cell biosensor that activates gene expression only under favorable conditions: in close proximity to plant roots and at high bacterial population densities. We engineered the pSal/nahR system in our host Pseudomonas fluorescens SBW25 to respond to salicylic acid, a key root exudate. In parallel, we implemented a quorum sensing system based on LuxI and the luxpR/LuxR pair to monitor cell density. Both inputs were integrated using a toehold switch-based AND gate, triggering expression only when both conditions were met. This strategy minimizes metabolic burden and offers a tightly controlled system for expression at target locations. While further validation in rhizosphere-like conditions is required, our results provide a foundation for safer open-environment applications of microorganisms, making this biosensor a versatile tool for future agricultural biotechnology.

Introduction: Preterm hypoxic-ischemic encephalopathy (pHIE) is a complex brain injury that contributes to chronic neural inflammation and neurological disorders. The signs and symptoms of in utero pHIE can often be overlooked, untreated or lumped into more generic conditions such as encephalopathy of prematurity (EOP). Clinical interventions like hypothermia and erythropoietin do not improve pHIE. We characterized a murine model for pHIE, which includes hypoxia and maternal factors as a cost-effective alternative to large animal models of HIE.

Methods: We injected pregnant mouse dams with LPS to stimulate an inflammatory response on embryonic days 15-16 (E15-E16), and whole cage hypoxia exposures occurred from postnatal days 3 to 9. To quantify the development of inflammation in the pHIE model, we used immunohistochemistry to stain for Caspase-9 in the cortex (20 μm per slice) and then counted Caspase-9 positive cells using unbiased stereology. We stained brain tissue with MAP2 to quantify neuronal intermediate filament expression and staining using a machine-learning based image analysis approach. We quantified cytokines (IL-1β, IL-6, IL-10, IL-18 and TNF-α) using RT-qPCR and (IL-18) ELISA to characterize differential expression in all treatment groups. The pHIE animals were compared with controls (LPS-Normoxia, Saline-Hypoxia, Saline-Normoxia, and Naïve) and with a model of only hypoxia (10% O₂) exposure in mouse pups.

Results: The pHIE pups showed significantly higher expression of Caspase-9 throughout the cortex compared to Naïve pup brains (p < 0.05). MAP2 expression was significantly decreased (p < 0.05) between 1.5-6.0 mm of the brain compared to Saline-Hypoxia and Naïve animals. Both IL-1β and IL-10 expression in LPS-Hypoxia animals was significantly higher (p < 0.05) than in Saline-Hypoxia and Naive animals. TNF-α expression was not significantly different between LPS-Hypoxia and Saline-Hypoxia animals. However, both showed significantly different transcription, compared to Naive animals.

Discussion: The model we describe here shows cortical damage similar to that seen in human HIE.