The investigation used two publicly available gene-expression datasets from the Gene Expression Omnibus. GSE26051 served as the training cohort, while differential-expression results from GSE180836 contributed to candidate-gene discovery and screening. Because GSE180836 informed candidate selection, it was not used as an independent validation cohort. The selected biomarkers were instead assessed by reverse-transcription quantitative PCR, or RT-qPCR, in a separate clinical cohort containing 30 participants. This design meant that the principal laboratory confirmation came from clinical samples rather than from a second independent analysis of the public transcriptomic datasets.

To identify differentially expressed genes associated with the macrophage polarization-related transcriptional signature, the researchers combined differential-expression analysis with weighted gene co-expression network analysis, known as WGCNA. Three machine-learning algorithms were then applied to screen candidate biomarkers: least absolute shrinkage and selection operator, support vector machine-based recursive feature elimination, and the Boruta algorithm. The analysis selected PTCHD1, NEURL1, and ST8SIA5, all of which were significantly downregulated in the training cohort GSE26051.

Clinical testing supported two of the three candidates more clearly than the third. PTCHD1 and NEURL1 were each significantly downregulated in tendinopathy samples relative to controls in the cohort of 30, with P values below 0.01 for both comparisons. ST8SIA5 also exhibited a downward trend, but the difference did not reach statistical significance, with P = 0.074. On this evidence, PTCHD1 and NEURL1 were the more strongly supported markers associated with tendinopathy, while ST8SIA5 remained a candidate showing a consistent but inconclusive expression pattern.

The researchers also constructed a diagnostic nomogram from the selected biomarkers and evaluated it using receiver operating characteristic and calibration analyses. The model achieved an area under the ROC curve above 0.7 and showed good calibration. These results indicate that the model discriminated between the groups represented in the study, but the reported evidence does not establish its diagnostic performance in larger or more diverse populations. The distinction is particularly important because the public dataset GSE180836 contributed to candidate discovery rather than serving as an independent validation set, while the clinical cohort used for RT-qPCR confirmation included only 30 participants.

Functional enrichment and gene-set analyses associated the biomarker-related network with pathways including PI3K–Akt signaling and extracellular matrix–receptor interaction. These are computational associations within the studied expression data, not evidence that PTCHD1 or NEURL1 directly controls macrophage polarization or that either gene is a therapeutic target. Similarly, a drug–gene interaction analysis identified 16 potential therapeutic compounds, but the source reports network-level candidates rather than experimental tests of treatment effects. The study’s strongest supported finding is therefore the clinical association of reduced PTCHD1 and NEURL1 expression with tendinopathy, alongside a gene-based model that showed discrimination above 0.7 AUC in the available validation material.

Subject of Research: Macrophage-associated transcriptional biomarkers for tendinopathy

Article Title: Identification of genes associated with a macrophage polarization-related transcriptional signature as biomarkers for tendinopathy based on bioinformatics analysis

Article References: Ren, L., He, C., & Qin, Y. (2026). Identification of genes associated with a macrophage polarization-related transcriptional signature as biomarkers for tendinopathy based on bioinformatics analysis. Scientific Reports. https://doi.org/10.1038/s41598-026-75576-z

Image Credits: AI Generated

DOI: 10.1038/s41598-026-75576-z

Keywords: tendinopathy, macrophage polarization, PTCHD1, NEURL1, ST8SIA5, gene biomarkers, RT-qPCR, WGCNA, machine learning, diagnostic nomogram, Identification, genes

The research team, led by Chengfeng Pan and colleagues at Cangnan Hospital of Wenzhou Medical University in China, took an unusually disciplined approach to a common pitfall in genomics. Rather than pooling datasets from different microarray platforms and hoping for the best, they analyzed four Gene Expression Omnibus datasets study by study, without any cross-platform batch harmonization. The discovery cohort, GSE59867, contained 111 admission samples from patients with ST-segment elevation myocardial infarction and 46 samples from patients with stable coronary artery disease. Three further cohorts, GSE48060, GSE60993, and GSE66360, totaling 348 participants across the study, served as fully independent validation sets. This design choice matters: by refusing to merge technically incompatible data, the researchers ensured that any signal surviving validation was biological rather than an artifact of preprocessing.

The first layer of analysis painted a vivid picture of the immune landscape in heart attack. Using marker-based scoring of nine immune cell programs, the team found that neutrophil and classical monocyte signatures were markedly elevated in STEMI patients, with Cliff’s delta effect sizes of 0.635 and 0.611 respectively and false discovery rates below 10^-8. In mirror image, natural killer cell and CD8 T-cell programs were suppressed. This myeloid surge with lymphoid retreat is exactly what cardiologists would predict from the biology of ischemia-reperfusion: dying heart muscle releases danger-associated molecular patterns that trigger neutrophil infiltration, monocyte extravasation, and macrophage activation. When that inflammatory wave clears debris and scaffolds repair, it is beneficial; when it persists, it drives fibrosis, maladaptive remodeling, and progressive cardiac dysfunction.

To connect genes to this immune picture, the researchers applied weighted gene co-expression network analysis, or WGCNA, to the 1,000 most variable genes in the discovery cohort. The network resolved into eight modules, and the largest, containing 215 genes, showed the strongest association with STEMI status, with a point-biserial correlation of 0.591 and an FDR of 2.97 x 10^-15. Gene ontology testing revealed that inflammatory response was the dominant biological theme within this module, with 20 of its 215 genes belonging to that pathway at an FDR of 2.41 x 10^-10. Hub topology analysis confirmed that genes most connected within the module were also the ones most strongly tied to disease status, a pattern that lends biological coherence to the statistical signal.

The centerpiece of the study, however, is the machine learning pipeline that distilled this complexity into a compact diagnostic index. The team used nested cross-validation, a design that strictly separates feature selection from performance estimation to prevent information leakage and overfitting. Within each of 25 outer test folds, genes were ranked by Welch-test p-value and the top eight selected; four algorithms, including L2-regularized logistic regression, linear support vector machines, random forest, and gradient boosting, then produced predictions. Three genes, BAX, EDA2R, and MCEMP1, were selected in every single outer fold, a striking consistency. The final fixed model settled on eight genes: BAX, EDA2R, MCEMP1, SOCS3, ASGR2, RNASE1, ADM, and KPNA5.

The performance numbers are eye-catching. In the discovery cohort, the repeated out-of-fold area under the receiver operating characteristic curve, a standard measure of diagnostic discrimination, reached 0.984 for logistic regression and 0.986 for the linear SVM, with a Brier score of 0.080 indicating good calibration. More importantly, the index transported to independent data: validation AUCs were 0.863 in GSE48060, 0.876 in GSE66360, and 0.655 in GSE60993. The authors are candid about the weaker result. GSE60993 mixed STEMI and NSTEMI patients, used whole blood rather than the PBMC or enriched cell preparations of the other cohorts, and contained only 24 samples, all factors that plausibly diluted the myeloid inflammatory signal the index was built to detect. A fourth dataset, GSE62646, yielded an AUC of 0.995 but was treated only as a technical sensitivity analysis because participant overlap with the discovery cohort could not be excluded.

What do the eight genes actually do? The study weaves them into a mechanistic narrative of sterile cardiac inflammation. BAX is a pro-apoptotic regulator of cell death, fitting for a disease defined by cardiomyocyte necrosis. EDA2R, which carried the largest positive coefficient in the final model with an odds ratio of 2.72 per standardized unit, is increasingly linked to aging and degenerative signaling. MCEMP1 marks mast cells and myeloid populations, while SOCS3 is a feedback inhibitor of cytokine signaling that spikes during inflammatory activation. ASGR2, RNASE1, ADM, and KPNA5 round out the panel with roles spanning hepatic clearance, extracellular RNA degradation, hypoxia adaptation, and nuclear transport. Notably, six of the eight genes retained the same direction of disease association across all five cohorts examined, though BAX and EDA2R reversed direction in two, a reminder that cross-cohort heterogeneity is real even when overall performance holds.

Pathway analysis added further depth. Among discovery genes ranked by test statistic, the strongest positive Hallmark signal was epithelial-mesenchymal transition, but IL-6/JAK/STAT3 signaling, inflammatory response, hypoxia, reactive oxygen species response, and TNF-alpha/NF-kB signaling all ranked positively as well, forming a connected network of leading-edge genes. A single-cell dataset, GSE121893, provided cellular context showing that the model genes display distinct expression patterns across endothelial cells, smooth-muscle cells, fibroblasts, and macrophages, with MCEMP1 concentrated in macrophages. The authors are careful to note that this dataset contained no actual AMI specimens, so it offers context rather than proof of AMI-specific localization.

The team also ran laboratory experiments to ground the computational findings in wet-lab reality. Using ATCC-authenticated THP-1 monocytes differentiated into macrophage-like cells and stimulated with lipopolysaccharide and ATP to activate the NLRP3 inflammasome, plus EA.hy926 endothelial cells subjected to hypoxia-reoxygenation, they measured four prespecified inflammatory transcripts by RT-qPCR. All four rose significantly: CXCL10 increased 3.82-fold, IL6 4.15-fold, NLRP3 2.94-fold, and CASP1 3.47-fold, each across three independent biological replicates. Because these transcripts were chosen as inflammasome-axis markers rather than members of the eight-gene model itself, the experiment corroborates the inflammatory program underlying AMI rather than directly validating the diagnostic index, a distinction the authors draw explicitly.

The study is equally forthright about its limits. Two exploratory outcome analyses, predicting six-month heart failure progression in 17 patients and 18-month recurrent events in 31, produced AUCs of 0.819 and 0.569 respectively, results the authors label descriptive only, given tiny event counts and the absence of patient-level censoring data. No Kaplan-Meier curves or hazard ratios were computed, and an exploratory two-subtype clustering of STEMI patients showed a numerical but non-significant difference in heart failure events. The researchers call for prospective validation with protocol-standardized blood sampling and blinded outcome adjudication before any clinical use. Still, the core achievement stands: a leakage-controlled, multi-cohort-validated blood-based gene index that captures the immune fingerprint of heart attack, built entirely on transparent, reproducible analysis of public data. If prospective studies confirm its accuracy, such a signature could one day complement troponin testing and help identify which patients carry the inflammatory profiles most likely to benefit from emerging immune-targeted therapies, from IL-6 receptor blockade to NLRP3 inhibitors now advancing through clinical development.

Subject of Research: A machine learning-derived eight-gene blood transcriptomic index for diagnosing acute myocardial infarction and characterizing its inflammatory immune response

Article Title: Cross-cohort blood transcriptome analysis identifies an eight-gene diagnostic index and inflammatory programs in acute myocardial infarction

Article References: Pan, C., Chen, Z., Li, J., Xie, S., & Fan, X. (2026). Cross-cohort blood transcriptome analysis identifies an eight-gene diagnostic index and inflammatory programs in acute myocardial infarction. Frontiers in Cardiovascular Medicine, 13, Article 1904246. https://doi.org/10.3389/fcvm.2026.1904246

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1904246

Keywords: acute myocardial infarction, blood transcriptomics, machine learning, diagnostic gene signature, inflammasome, NLRP3, immune response, WGCNA, biomarker validation, precision cardiology, CXCL10, heart failure

The research team, led by investigators affiliated with West China Hospital of Sichuan University, recruited 49 patients with first-episode, drug-naïve major depressive disorder and 35 healthy controls. Restricting the comparison to people who had never received antidepressant medication was a deliberate methodological choice, because drugs such as selective serotonin reuptake inhibitors are known to alter both gut microbial composition and epigenetic marks. By studying participants before any pharmacological intervention, the researchers aimed to capture biological signals more closely tied to the illness itself rather than to its treatment.

To quantify circadian disruption, the team used the Munich Chronotype Questionnaire, a validated self-report instrument that characterizes individual differences in sleep timing. Two derived metrics anchored the analysis: corrected mid-sleep time on free days, which estimates a person’s internal sleep phase once the constraints of workdays are removed, and social jetlag, which measures the mismatch between that internal timing and the schedule imposed by social obligations. Social jetlag has attracted increasing attention in chronobiology because it captures a form of chronic circadian misalignment that is common in modern life and has been linked to metabolic and mood disturbances in prior research.

The microbial arm of the study relied on shotgun metagenomic sequencing of fecal samples, a technique that reads the collective genomes of gut bacteria directly rather than profiling only a marker gene. This approach allows researchers to characterize microbial communities at higher resolution than targeted amplicon sequencing. Meanwhile, the epigenetic arm used reduced-representation bisulfite sequencing on peripheral blood DNA, a method that enriches for regions dense in CpG sites, the cytosine-guanine sequences where methylation changes most often influence gene regulation. Bisulfite treatment converts unmethylated cytosines into uracil, allowing methylation status at individual sites to be read out across the genome.

The first major finding concerned community structure rather than richness. Alpha diversity, which measures the number and evenness of microbial species within a single individual, was comparable between patients and controls. Beta diversity, by contrast, which quantifies differences in community composition between individuals, revealed significantly distinct gut microbial communities in the depressed group. In practical terms, patients did not necessarily carry fewer microbial species, but the identity and relative abundance of those species differed in ways that separated them from healthy participants.

To make sense of the high-dimensional microbial data, the researchers applied weighted gene co-expression network analysis, a computational framework originally developed for genomics that groups correlated features into modules and tests whether module-level behavior tracks clinical variables. When applied to the metagenomic profiles, this approach detected microbial modules whose abundance patterns were significantly correlated with the circadian parameters, suggesting that the relationship between sleep timing and gut ecology is not scattered across individual taxa but organized into coherent community-level signatures.

The methylation analysis yielded an even larger signal. Reduced-representation bisulfite sequencing identified 9,749 differentially methylated genes in the depression group, and gene ontology and KEGG pathway analyses showed that these genes were enriched in neural-related pathways, including processes tied to neurodevelopment and synaptic function. The authors are careful to frame this appropriately: peripheral blood methylation does not directly mirror methylation in brain tissue, and the enrichment patterns provide hypothesis-generating evidence for systemic associations rather than proof that the observed epigenetic changes reflect what is happening in the central nervous system.

The most striking result emerged when the three data layers were integrated. Correlation analyses linking the microbial modules, the circadian metrics, and the methylation data identified nine key metagenomic operational taxonomic units that were simultaneously associated with circadian rhythm measures and with the methylation rates of 898 differentially methylated genes related to neurodevelopment and synaptic function. This tripartite overlap is what elevates the study beyond parallel descriptions of three altered systems. It proposes a specific, testable network in which sleep timing, particular gut microbes, and the epigenetic regulation of neural genes are interconnected in depression.

Several biological mechanisms could plausibly underlie such a network. Gut microbes produce and modulate short-chain fatty acids, metabolites that influence both host metabolism and epigenetic machinery, since butyrate in particular is known to inhibit histone deacetylases and thereby affect chromatin state. Circadian disruption can alter gut microbial rhythms through feeding time and stress hormone fluctuations, and the hypothalamic-pituitary-adrenal axis provides a well-documented channel through which circadian and microbial signals can influence systemic physiology, including immune and epigenetic processes. The study was not designed to test these mechanisms directly, but the observed correlations are consistent with such frameworks.

The authors and the data itself are explicit about the study’s limits. It is a cross-sectional pilot with a modest sample size, meaning it captures associations at a single point in time and cannot establish whether circadian disruption drives microbial change, whether microbial metabolites shape methylation, whether methylation influences mood symptoms, or whether all three reflect a shared upstream cause. The findings are best read as a preliminary multi-omic network hypothesis, a map of candidate relationships that now invites longitudinal follow-up, mechanistic experiments in model systems, and replication in larger and more diverse cohorts. If the network holds up, it could point toward interventions that target sleep timing and gut ecology as complementary levers in depression care.

Subject of Research: Multi-omics associations among circadian rhythm disruption, gut microbiota, and DNA methylation in major depressive disorder

Article Title: Associations between circadian rhythm disruption, specific gut microbiota, and DNA methylation in major depressive disorder: a pilot study

Article References: Fan, H., Wang, Y., Guan, X., Yuan, Y., Xie, Q., & Dong, Z. (2026). Associations between circadian rhythm disruption, specific gut microbiota, and DNA methylation in major depressive disorder: a pilot study. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08619-4

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08619-4

Keywords: major depressive disorder, circadian rhythm, gut microbiota, DNA methylation, social jetlag, shotgun metagenomics, reduced-representation bisulfite sequencing, WGCNA, multi-omics, epigenetics, psychiatry, neurodevelopment

The study utilized data from The Cancer Genome Atlas Breast Cancer (TCGA-BRCA) cohort, which included 1,113 tumor samples and 113 normal breast tissue samples. To identify genes whose expression patterns are significantly altered during tumor development, the researchers employed weighted gene co-expression network analysis (WGCNA). This computational technique constructs a network where genes are nodes and their co-expression correlations are edges, allowing for the identification of modules of highly interconnected genes. By using normal breast tissue as a reference network, the team assessed how these modules were preserved or disrupted across different stages of breast cancer, from stage I to stage IV.

Among the ten co-expression modules identified, the turquoise module demonstrated the weakest preservation pattern across tumor-derived networks, indicating that the genes within this module undergo significant regulatory changes during cancer progression. Functional enrichment analysis revealed that this module is heavily involved in epithelial development, cell adhesion, and extracellular matrix interactions. Key pathways included ECM-receptor interaction and the PI3K-Akt signaling pathway, both of which are critical in tumor progression and metastasis. The disruption of these networks suggests a loss of normal tissue architecture and communication, a hallmark of aggressive malignancy.

To pinpoint specific biomarkers, the researchers constructed a protein-protein interaction network from the genes in the least-preserved module and used eleven different algorithms to rank hub genes. From 120 candidate genes, six were selected based on their independent association with overall survival in multivariable Cox regression analysis: CLDN9, SDC1, IL2, CLIC6, KRT5, and MMP25. These genes were used to construct a risk score model, where each gene’s expression level is weighted by its regression coefficient. Patients with higher risk scores, driven largely by the expression of CLDN9 and SDC1, exhibited significantly poorer survival outcomes compared to those with lower scores.

The prognostic performance of the six-gene signature was rigorously validated in independent external cohorts, including three datasets from the Gene Expression Omnibus (GSE20685, GSE42568, and GSE58812) and the large METABRIC cohort. The model demonstrated consistent discriminative ability across these diverse populations, with area under the curve (AUC) values ranging from approximately 0.53 to 0.72 depending on the time point and cohort. In the TCGA training set, the model achieved AUC values of 0.641, 0.723, and 0.689 for 1-, 3-, and 5-year overall survival, respectively. The robustness of the model was further confirmed by its ability to predict disease-specific survival, disease-free interval, and progression-free interval, indicating that the signature captures broad aspects of tumor aggressiveness.

Beyond survival prediction, the study investigated the relationship between the risk score and the tumor immune microenvironment. Single-sample Gene Set Enrichment Analysis (ssGSEA) and the ESTIMATE algorithm revealed that patients in the low-risk group exhibited a more immune-enriched profile. Specifically, the low-risk group showed higher infiltration of T cells, natural killer cells, and antigen-presenting cells, along with elevated scores for immune-related functional signatures such as antigen presentation and cytotoxic activity. In contrast, the high-risk group was characterized by a more immunosuppressive microenvironment, with reduced immune cell infiltration and higher expression of proliferation-associated pathways such as E2F targets and G2M checkpoint.

The distinct immune phenotypes observed between the risk groups prompted an evaluation of potential immunotherapy response using computational tools such as the Immunophenoscore (IPS) and Tumor Immune Dysfunction and Exclusion (TIDE) algorithms. The low-risk group displayed higher IPS values, suggesting a more immunogenic phenotype and potentially greater sensitivity to immune checkpoint blockade. However, TIDE analysis presented a more complex picture, with the low-risk group showing higher dysfunction scores, which might indicate a less favorable predicted response in some contexts. This discordance between IPS and TIDE scores highlights the complexity of the tumor immune landscape and suggests that the six-gene signature captures distinct immune-related characteristics that may not uniformly predict immunotherapy benefit.

Further analysis of the individual genes revealed their diverse biological roles in cancer. CLDN9, a tight junction protein, has been linked to tumor invasion and metastasis in other cancers, while SDC1, a transmembrane proteoglycan, is associated with tumor progression and brain metastasis in breast cancer. IL2, a key cytokine in immune regulation, showed favorable prognostic associations, consistent with its role in T-cell proliferation. KRT5, a keratin protein, and MMP25, a matrix metalloproteinase, were also found to be significantly associated with survival, reflecting the interplay between epithelial structure, immune signaling, and extracellular matrix remodeling in breast cancer biology.

The study also evaluated the diagnostic potential of the six-gene signature in an independent microarray cohort (GSE86374). The combined logistic regression model achieved an AUC of 0.927 for distinguishing breast tumor tissues from normal tissues, demonstrating strong diagnostic accuracy. Individual genes such as SDC1 and KRT5 showed significant expression differences between tumor and normal samples, further supporting their utility as biomarkers. The integration of these genes into a single score provides a comprehensive molecular fingerprint that reflects both the intrinsic biology of the tumor and its interaction with the host immune system.

While the six-gene signature shows promise for risk stratification and molecular characterization, the authors note that prospective validation in larger multicenter cohorts is necessary to confirm its clinical applicability. Additionally, functional studies are required to elucidate the precise mechanisms by which these genes influence tumor progression and immune evasion. Nevertheless, this work represents a significant step toward integrating network-based bioinformatics with clinical outcomes, offering a potential tool for personalized medicine in breast cancer. By linking gene expression patterns to immune microenvironment characteristics, the study provides new insights into the heterogeneity of breast cancer and the factors that drive patient outcomes.

Subject of Research: Development and validation of a six-gene prognostic signature associated with tumor microenvironment characteristics in breast cancer

Article Title: Identification and validation of a six-gene prognostic signature associated with tumor microenvironment characteristics in breast cancer

Article References: Keshavarzi, F., Akhoundi, F., & Hosseini, A. (2026). Identification and validation of a six-gene prognostic signature associated with tumor microenvironment characteristics in breast cancer. Heliyon, 12(16), Article e45601. https://doi.org/10.1016/j.heliyon.2026.e45601

Image Credits: AI Generated

DOI: 10.1016/j.heliyon.2026.e45601

Keywords: breast cancer, prognostic signature, tumor microenvironment, WGCNA, gene expression, immunotherapy, bioinformatics, survival analysis, Identification, validation, six-gene, prognostic

The research team, led by Hong Liu, Zicheng Liang, Jun Tan, and Chunyun Yuan from institutions in Changsha and Hangzhou, China, took an unusual approach to a familiar problem. Rather than running new laboratory experiments, they mined a publicly available dataset of skeletal muscle biopsies from ten subjects who underwent three distinct phases: an initial baseline, nine days of strict bed rest that reliably induces insulin resistance, and four weeks of exercise retraining that reverses it. This bidirectional design, in which the same individuals pass from normal metabolism into insulin resistance and back again, offered a rare opportunity to identify genes whose expression tracks the reversible dynamics of insulin sensitivity rather than static disease states.

To sift through the transcriptomic data, the researchers combined two complementary strategies. Weighted gene co-expression network analysis, or WGCNA, grouped thousands of genes into modules whose collective activity correlated with the metabolic phase, while differential expression analysis flagged individual genes whose levels changed significantly between phases. In the bed rest comparison, a module of 319 genes correlated strongly with the insulin-resistant state, and 66 genes were differentially expressed. In the exercise retraining comparison, two modules containing 237 and 417 genes tracked recovery, with 46 differentially expressed genes. Intersecting these results produced candidate gene pools of 46 and 42 genes for the two phases respectively.

The critical filtering step came from machine learning. The team applied LASSO regression, a technique that shrinks the coefficients of less informative genes to zero, and random forest analysis, which ranks genes by how much they improve classification accuracy across thousands of decision trees. Only genes that survived both algorithms were considered candidate targets. When the dust settled, WNT4 emerged as the sole gene selected by both methods in both the bed rest and exercise retraining phases. Its expression followed a striking bidirectional pattern: it rose after bed rest and fell again after exercise retraining, suggesting that its levels are tightly coupled to the metabolic status of the muscle rather than reflecting a one-way injury response.

To assess how reliably WNT4 could distinguish between metabolic states, the researchers used leave-one-out cross-validation, a rigorous scheme in which each sample is predicted by a model trained on all the others. This approach avoids the inflated accuracy estimates that plague small-sample studies using conventional neural networks. WNT4 achieved an area under the receiver operating characteristic curve of 0.88 in the bed rest comparison and 0.87 in the exercise retraining comparison, remarkably consistent values across two independent physiological transitions. Other candidate genes, including AQP7, GPD2, CTGF, and FOXM1, performed well in one phase but not both, underscoring WNT4’s unique position.

Gene set enrichment analysis then revealed what WNT4 might actually be doing. When samples were split into high- and low-WNT4-expression groups, the pathways positively correlated with high expression were dominated by energy metabolism, including oxidative phosphorylation, the TCA cycle, fatty acid metabolism, and the PPAR signaling pathway. This focused functional signature led the authors to propose what they call an endogenous compensation hypothesis: during the early phase of disuse, WNT4 may act as a protective factor that helps muscle stabilize its mitochondrial network and maintain metabolic homeostasis. WNT4 is known to inhibit the canonical Wnt signaling pathway and to keep muscle stem cells in a quiescent state, both of which could serve protective roles under metabolic stress.

But the hypothesis carries a darker corollary. If bed rest stress persists, the authors speculate, sustained WNT4 overexpression might excessively suppress the canonical Wnt and beta-catenin pathway, which previous studies have shown is needed for normal insulin-stimulated glucose uptake in skeletal muscle. In this scenario, the initial compensatory response would transition into a maladaptive state, contributing to intramyocellular lipid deposition and impaired insulin signaling through nodes such as AMPK and IRS-1. The authors describe this as a stage transition from endogenous compensation to pathological decompensation, a framework they acknowledge is entirely hypothetical and requires experimental validation through WNT4 knockdown and overexpression studies.

The regulatory architecture behind WNT4 adds another layer of complexity. Using database predictions, the team constructed a network linking transcription factors, messenger RNAs, microRNAs, and long non-coding RNAs. The transcription factor EGR1 emerged as the predicted upstream activator of WNT4 transcription, while the microRNA hsa-miR-1238-3p, buffered by the long non-coding RNAs LINC00969 and PRKCQ-AS1, was predicted to fine-tune WNT4 expression at the post-transcriptional level. Two circular regulatory loops involving these components scored above 0.90 on both random forest and support vector machine confidence metrics. The authors stress that these relationships are purely computational predictions and have not been experimentally confirmed.

Perhaps the most ambitious part of the study addresses a translational bottleneck: a search of the DrugBank database confirmed that no approved drugs currently target WNT4 directly. To fill this gap, the researchers built a computational drug discovery pipeline. They started with the amino acid sequence of human WNT4 and used AlphaFold to predict its three-dimensional structure, achieving an average confidence score of 93.12 on the pLDDT scale. Because no experimental crystal or cryo-EM structure of human WNT4 exists, the predicted binding pocket, located near residues CYS208, GLY210, and SER214 around the known lipid modification site at Ser212, remains a computational construct awaiting validation.

From there, the team deployed a deep learning docking model called DiffDock, which learns the conformational distribution of protein-ligand binding from data rather than relying on fixed sampling rules. After filtering more than 700,000 compounds from natural product and marine product databases using Lipinski’s rules and a quantitative drug-likeness score, 79,319 molecules were successfully docked against the predicted WNT4 pocket. Three compounds emerged at the top: CMNPD4923, identified as the marine sponge alkaloid Damirone B; CMNPD27319, the diaryl ether Aspergilol E from marine Aspergillus fungi; and CMNPD30941, a substituted isochromanone derivative. All three formed hydrogen bonds with the same pocket residues at distances between 1.70 and 2.42 angstroms.

One hundred nanosecond molecular dynamics simulations confirmed that all three protein-ligand complexes reached stable equilibrium, with root mean square deviations settling between 0.67 and 0.86 nanometers during the second half of the simulations. Binding free energy calculations using the MM-GBSA method ranked CMNPD4923 strongest at minus 20.08 kilocalories per mole, followed by CMNPD27319 at minus 17.13, while CMNPD30941 lagged at minus 5.80. Notably, this ranking diverged from the DiffDock confidence scores, suggesting that no single computational metric is sufficient and that multi-scale validation is essential. Density functional theory calculations added a quantum mechanical layer, revealing that the compounds possess active frontier orbital distributions with potential for charge transfer, though these gas-phase calculations approximate rather than replicate the protein pocket environment.

The authors are careful to frame every result as a hypothesis-generating prediction rather than a confirmed finding. The sample size of ten subjects is small, WNT4 was not cross-validated in an independent cohort, the protein structure is predicted rather than experimentally resolved, and the binding energies omit entropy contributions. Actual binding affinity would need to be confirmed by surface plasmon resonance or microscale thermophoresis, and functional effects by cellular assays. Nevertheless, the study demonstrates a replicable computational framework that spans from classical mechanics to quantum chemistry, and it offers a concrete chemical starting point for a target that has never been drugged. If future experiments validate the endogenous compensation hypothesis, WNT4 could become a new focal point for understanding how exercise reprograms muscle metabolism, and the three marine-derived compounds could serve as lead structures for developing the first targeted therapeutics aimed at this enigmatic signaling molecule.

Subject of Research: Computational identification of WNT4 as a biomarker and drug target in exercise-mediated skeletal muscle insulin sensitivity

Article Title: Computational identification of candidate biomarkers and compounds associated with skeletal muscle insulin sensitivity improvement mediated by physical activity: an integrative multi-omics and computational chemistry study

Article References: Liu, H., Liang, Z., Tan, J., & Yuan, C. (2026). Computational identification of candidate biomarkers and compounds associated with skeletal muscle insulin sensitivity improvement mediated by physical activity: an integrative multi-omics and computational chemistry study. Frontiers in Bioinformatics, 6, Article 1885995. https://doi.org/10.3389/fbinf.2026.1885995

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1885995

Keywords: WNT4, insulin sensitivity, skeletal muscle, physical activity, insulin resistance, machine learning, WGCNA, molecular docking, DiffDock, molecular dynamics, natural products, drug discovery

The epidemiological starting point came from the Global Burden of Disease 2021 study, which the team mined for data on periodontal diseases among people aged five to twenty from 1999 to 2021. The burden remained substantial throughout the period, with only modest temporal change overall, but the analysis revealed striking heterogeneity across sociodemographic strata. Low- and low-middle sociodemographic index regions consistently showed higher age-standardized prevalence, incidence, and disability-adjusted life-year rates than high-SDI regions, and male adolescents generally carried a slightly heavier burden than females. The authors stress that because PFOS is not a modeled GBD risk factor for periodontal disease, these figures provide descriptive public health context rather than direct evidence of chemical causation. What they do establish is a persistent socioeconomic inequity in adolescent gum disease that motivates the search for modifiable environmental contributors.

That search turned to Mendelian randomization, a genetic technique that uses naturally occurring DNA variants as instrumental variables to test whether an exposure plausibly causes a disease. Using 65 genetic instruments associated with circulating PFOS levels, drawn from more than 8,000 participants in the Canadian Longitudinal Study on Aging, and chronic periodontitis summary statistics from over 195,000 FinnGen participants of European ancestry, the team applied the inverse-variance weighted method as its primary estimator. The result was a statistically significant positive association: genetically predicted PFOS exposure increased periodontitis risk with an odds ratio of 1.148 (95 percent confidence interval 1.053 to 1.251, p = 0.0017). Complementary MR-Egger, weighted-median, and mode-based estimates pointed in the same direction, though several did not reach conventional significance, a pattern the authors attribute to limited statistical power rather than conflicting evidence.

Sensitivity analyses were designed to rule out the artifacts that plague observational studies. Cochran’s Q statistics detected no substantial heterogeneity among instruments, the MR-Egger intercept and MR-PRESSO global tests found no directional horizontal pleiotropy, the funnel plot was largely symmetric, and leave-one-out analyses showed that removing any single variant did not materially change the estimate. The authors are careful about the limits of the design: the analysis relied on a single PFOS GWAS without independent replication or colocalization, the datasets were predominantly European-ancestry and non-adolescent, and linkage disequilibrium confounding cannot be fully excluded. They therefore present the genetic evidence as supportive rather than definitive, while noting it is far stronger than a simple cross-sectional correlation between chemical levels and gum disease.

To uncover the molecular machinery that might mediate such an effect, the researchers merged two public periodontitis transcriptomic datasets, GSE16134 and GSE173078, correcting batch effects with ComBat and surrogate variable analysis. Differential expression analysis identified 1,005 genes altered between diseased and healthy gingival tissue, and weighted gene co-expression network analysis pinpointed disease-associated modules. Intersecting these lists yielded 329 candidate periodontitis-related genes, which the team then overlapped with a curated set of 8,827 PFOS-related human proteins compiled from the Comparative Toxicogenomics Database and prediction tools. The intersection produced 151 putative PFOS-periodontitis targets. Enrichment analyses of these genes converged on blood coagulation, leukocyte migration, phagocytosis, extracellular matrix organization, serine-type peptidase activity, and canonical inflammatory cascades including JAK-STAT, MAPK, TNF, IL-17, and NF-κB signaling, precisely the pathways that govern the destructive host immune response in periodontal tissue.

From the 151 candidates, a machine-learning framework that evaluated 127 model configurations across lasso, support-vector machine, random forest, XGBoost, and other algorithms, with stratified five-fold cross-validation and two independent validation cohorts, distilled a ten-gene biomarker panel: APLNR, AQP9, BTG2, CXCL1, HBD, MME, MMP13, SERPINB9, STK17B, and XBP1. Each gene showed good individual diagnostic ability, with area under the ROC curve exceeding 0.78, and a multigene nomogram built from the panel achieved an AUC of 0.939 with good calibration and favorable decision-curve net benefit. SHAP explainability analysis quantified each gene’s contribution and revealed nonlinear relationships between expression and predicted risk. Many of the genes carry mechanistic plausibility: CXCL1 attracts neutrophils, MMP13 degrades collagen, XBP1 governs endoplasmic reticulum stress and immune function, and STK17B participates in lymphocyte signaling.

Immune deconvolution added a cellular dimension. Applying CIBERSORT to the bulk transcriptomic samples, the team inferred the relative abundance of 22 immune cell subsets and found marked shifts in periodontitis, particularly among macrophages, neutrophils, dendritic cells, T cell subsets, and mast cells. Several core genes correlated positively with pro-inflammatory populations such as M1 macrophages, neutrophils, and activated CD4-positive T cells, and negatively with resting or naive subsets. Single-cell RNA sequencing of peripheral blood from eleven periodontitis patients and ten controls then showed that eight of the detectable core genes were enriched in myeloid and T cell clusters. Ligand-receptor analysis revealed a dense communication network in which macrophages and T cells acted as major hubs, with macrophage migration inhibitory factor, or MIF, emerging as a prominent signaling axis connecting myeloid cells and T cells. Because the single-cell data came from blood rather than gingival lesions, the authors interpret this as evidence of systemic immune dysregulation, and the PFOS-MIF link remains hypothesis-generating.

Structural biology provided a physical rationale. Molecular docking with AutoDock Vina showed that PFOS binds favorably to all ten target proteins, with the strongest predicted affinity for STK17B at minus 10.4 kilocalories per mole. The binding mode was consistent across targets: the chemical’s sulfonate head group forms polar contacts with nearby residues while its perfluoroalkyl tail inserts into a hydrophobic pocket. A 100-nanosecond molecular dynamics simulation in GROMACS confirmed that the PFOS-STK17B complex is conformationally stable, with root-mean-square deviation settling between 0.35 and 0.55 nanometers, a steady radius of gyration, two to four persistent intermolecular hydrogen bonds, and a dominant low-energy basin in the free-energy landscape. The authors caution that docking and simulation establish structural plausibility, not functional proof.

Experimental validation brought the computational chain back to the bench. Human gingival fibroblasts exposed to 64 micromolar PFOS for 24 hours, a concentration selected after a CCK-8 viability assay showed marked toxicity only at higher doses, showed significantly elevated mRNA expression of all ten core genes by RT-qPCR. The authors frame the full picture honestly: the adolescent GBD data offer context, the Mendelian randomization offers quasi-causal genetic evidence in a broader population, and the omics, structural, and in vitro results offer mechanisms and biomarkers, but the datasets were not adolescent-specific, clinical covariates were unavailable, and dose-response relationships remain untested. Still, the convergence is notable. With forever chemicals under intensifying regulatory scrutiny and adolescent gum disease concentrated where dental resources are scarcest, the study argues that PFOS deserves evaluation as a potentially modifiable environmental risk factor for periodontal health, and its ten-gene signature may serve as an early-warning biomarker for environmentally associated periodontal destruction.

Subject of Research: PFOS exposure and its genetic, immune, and molecular links to periodontitis risk

Article Title: PFOS exposure and periodontitis from adolescent burden to genetic and multiomics evidence

Article References: Wan, H., Wang, Z., Peng, H., Wang, X., Xiong, L., & Zou, X. (2026). PFOS exposure and periodontitis from adolescent burden to genetic and multiomics evidence. iScience, 29(11), Article 117905. https://doi.org/10.1016/j.isci.2026.117905

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117905

Keywords: PFOS, PFAS, periodontitis, Mendelian randomization, forever chemicals, gingival fibroblasts, WGCNA, machine learning biomarkers, molecular docking, immune infiltration, Global Burden of Disease, adolescent health

The study, published in Frontiers in Bioinformatics by Yumeng Liu and Gang Luo of the Department of Pediatrics at Liaoning Maternal and Child Health Hospital, took aim at a long-standing gap in sepsis medicine. Childhood sepsis cannot be regarded as a mere miniature version of adult sepsis. Children undergo continuous development of their immune systems, and the inflammatory response, alterations in organ function, and clinical diagnostic criteria following infection all exhibit age-related characteristics. Consequently, candidate markers identified in adult cohorts cannot be directly applied to the diagnostic context observed in children. By screening and validating molecular markers specifically within child cohorts, the researchers sought diagnostic insights more aligned with the clinical and biological characteristics of pediatric sepsis.

The scientific premise of the work rests on oxidative stress, a state that arises when reactive oxygen species, or ROS, are produced in amounts greater than the body can neutralize with antioxidants. This imbalance drives mitochondrial dysfunction, which has been identified as a central pathophysiological mechanism underlying organ failure in sepsis, both in experimental models and in clinical patients. In a landmark study of 28 severe sepsis patients during the first 24 hours of admission, Brealey and colleagues observed that mitochondrial dysfunction, along with antioxidant depletion and impaired energy metabolism, were key drivers of multiple organ failure. Recent clinical studies in pediatric populations have demonstrated that affected children exhibit significant alterations in oxidative stress markers and antioxidant capacity, indicating that disruption of the oxidative-antioxidant balance may play a critical role in the pathophysiology of pediatric sepsis. Excessive ROS generation also triggers the MAPK and PI3K/Akt signaling pathways, elevates NF-κB levels, and consequently induces the expression of pro-inflammatory genes, and elevated NF-κB in sepsis patients serves as a stronger predictor of disease severity than conventional inflammatory markers.

To hunt down the molecular fingerprints of this oxidative storm, the researchers mined five open-access transcriptomic datasets from the Gene Expression Omnibus database, all comprising peripheral whole-blood samples from children with sepsis and healthy controls, all generated on the same GPL570 microarray platform to ensure comparability. Two datasets, GSE26440 and GSE26378, were merged to form the training cohort, while three independent datasets, GSE13904, GSE9692, and GSE8121, were combined into a validation cohort. The team corrected batch effects using the empirical Bayes ComBat method, verified the integration with principal component analysis, and then applied differential expression analysis with the limma package, using thresholds of |log2 fold change| greater than 0.585 and an adjusted p-value below 0.05. This rigorous pipeline yielded 396 significant differentially expressed genes, including 290 upregulated and 106 downregulated genes, and after intersecting these with a curated list of 817 oxidative stress-associated genes from prior literature, 36 differentially expressed oxidative stress-related genes emerged as the study’s raw material.

Functional enrichment analysis revealed what these 36 genes were doing. Gene Ontology analysis highlighted their roles in regulating inflammatory responses and reactive oxygen species metabolism, while Kyoto Encyclopedia of Genes and Genomes pathway analysis identified nine pathways implicated in sepsis progression, including neutrophil extracellular trap formation, the Toll-like receptor and NOD-like receptor signaling pathways that orchestrate innate immunity, and the regulation of inflammatory mediators affecting TRP channels. Neutrophil extracellular traps, or NETs, the DNA-based web-like structures that neutrophils deploy to capture and kill pathogens, were flagged as particularly critical in sepsis pathogenesis. NET formation relies on ROS generated by NADPH oxidase, and in sepsis an excessive immune response drives uncontrolled NET production that aggravates lung injury. Gene set enrichment analysis further showed that the oxidative stress phenotypic score was significantly higher in the sepsis cohort than in controls, with upregulation of ROS metabolic processes and sepsis-associated pathways in the high-score group.

The methodological heart of the study was a two-stage computational gauntlet. First, weighted gene co-expression network analysis, or WGCNA, was applied to the top 15 percent most variable genes in the training cohort. At a soft-thresholding power of beta equal to 8, the network closely approximated a scale-free topology, and the dynamic tree-cut algorithm resolved 14 distinct co-expression modules. One module, dubbed turquoise, showed a striking positive correlation with sepsis, with a Spearman coefficient of 0.65 and a p-value of 4.57 times 10 to the power of minus 29, marking its member genes as collectively pivotal in disease progression. Second, six machine learning algorithms, Bayesian analysis, bagged trees, LASSO regression, random forest, learning vector quantization, and Boruta, were each set loose on the oxidative stress-related differentially expressed genes. Only genes identified by all six methods, a threshold of six out of six, were retained as robust candidates. This consensus rule ensured that the screening results were not dependent on any single algorithm, and it ultimately distilled the field down to seven candidate diagnostic genes: DDAH2, GADD45A, HP, MAPK14, PGD, RETN, and S100A9.

From these seven candidates, the researchers constructed a protein-protein interaction network using the STRING database and excluded DDAH2 for low connectivity, leaving six hub genes: GADD45A, HP, MAPK14, PGD, RETN, and S100A9. Each of these genes encodes a protein with a plausible mechanistic link to sepsis biology. MAPK14, better known as p38 alpha, is a central kinase in inflammatory signaling cascades. S100A9 is an alarmin released abundantly by activated neutrophils and monocytes. HP is haptoglobin, an acute-phase protein that binds free hemoglobin. PGD encodes an enzyme in the pentose phosphate pathway central to cellular antioxidant defense. GADD45A participates in DNA repair and stress responses, and RETN, or resistin, is an inflammatory adipokine. Expression levels of all six hub genes were significantly elevated in septic children compared with controls, and receiver operating characteristic analysis revealed area under the curve, or AUC, values of 0.951, 0.982, 0.964, 0.944, 0.962, and 0.982 in the training data, and between 0.908 and 0.977 in the independent validation cohort, a level of diagnostic performance that few single biomarkers in clinical use today can match.

The team did not stop at individual genes. Seven machine learning classifiers, including random forest, support vector machine, LogitBoost, k-nearest neighbors, Naive Bayes, AdaBoost, and CancerClass, were trained on the pooled training cohort with five-fold cross-validation repeated ten times, and then evaluated on the untouched validation cohort. The k-nearest neighbors model, which classifies samples by their similarity to neighboring expression profiles, achieved the highest AUC in validation and was selected as the final predictive model. A nomogram built from the six hub genes pushed diagnostic accuracy even further, reaching an AUC of 0.994. The researchers also performed immune cell deconvolution using CIBERSORT, estimating the relative proportions of 22 immune cell types in peripheral blood. The analysis revealed marked alterations in the sepsis cohort: M0 macrophages, M2 macrophages, monocytes, activated mast cells, and neutrophils were significantly elevated, while follicular helper T cells, CD8 T cells, CD4 memory T cells, resting dendritic cells, and activated NK cells were markedly reduced. Most hub genes were positively correlated with neutrophils and negatively linked with resting dendritic cells, suggesting that oxidative stress-related genes may modulate the abundance and activity of distinct immune cell subsets in the immune microenvironment.

Perhaps the most clinically tantalizing result came from the drug-discovery arm of the study. The researchers queried the Drug-Gene Interaction database and the Connectivity Map to identify small molecules predicted to downregulate the hub genes, then performed blind molecular docking using the CB-Dock2 platform. The MAPK14 protein demonstrated strong binding affinities for doxycycline and ralimetinib mesylate, with AutoDock Vina scores of minus 8.4 and minus 8.1 kilocalories per mole respectively, while the S100A9 protein exhibited significant binding affinity for tasquinimod at minus 7.0 kilocalories per mole. Doxycycline, an antibiotic already in widespread clinical use, engaged MAPK14 through multiple ionic bonds, hydrophobic forces, and hydrogen bonds, while ralimetinib mesylate, a p38 MAPK inhibitor, and tasquinimod, an experimental immunomodulator, engaged their targets through complementary interaction networks. These findings position MAPK14 and S100A9 as promising therapeutic targets and open up new treatment options that could be explored for repurposing in pediatric sepsis.

The authors are careful to frame their findings as a foundation rather than a finished clinical tool. The analysis is based entirely on publicly available retrospective datasets, and heterogeneity across cohorts in sample sources, platforms, and clinical annotations may still influence performance estimates. The immune cell deconvolution relied on a single algorithm and reflects peripheral blood composition rather than tissue-level infiltration, and the molecular docking results suggest potential binding but do not demonstrate therapeutic efficacy in patients. The ceRNA network of 102 nodes and 160 edges, which implicates 89 long non-coding RNAs and 10 microRNAs in the post-transcriptional regulation of GADD45A, MAPK14, and PGD, remains exploratory pending experimental validation. Still, the study’s significance is hard to overstate. Compared with conventional biomarkers such as procalcitonin, C-reactive protein, and lactate, this six-gene oxidative stress signature may more sensitively reflect early molecular alterations associated with immune dysregulation, providing complementary diagnostic information precisely when conventional indicators are inconclusive. If prospective, multicenter validation confirms these results, the era of catching pediatric sepsis in its first critical hours, before organ failure takes hold, may move one decisive step closer to reality.

Subject of Research: Oxidative stress-related biomarkers and machine learning diagnostic models for pediatric sepsis

Article Title: Identification and validation of biomarkers related to oxidative stress in pediatric sepsis utilizing WGCNA and multiple machine learning algorithms

Article References: Liu, Y., & Luo, G. (2026). Identification and validation of biomarkers related to oxidative stress in pediatric sepsis utilizing WGCNA and multiple machine learning algorithms. Frontiers in Bioinformatics, 6, Article 1745381. https://doi.org/10.3389/fbinf.2026.1745381

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1745381

Keywords: pediatric sepsis, oxidative stress, biomarkers, machine learning, WGCNA, GADD45A, MAPK14, S100A9, molecular docking, immune cell deconvolution, diagnostic model, reactive oxygen species

Rare earth elements are a group of seventeen chemically similar metals, including neodymium, cerium, and ytterbium, that have become indispensable to modern technology. They appear in magnets, batteries, catalysts, and electronic displays, and as mining and manufacturing have intensified worldwide, these elements have quietly accumulated in soil, water, and air. People are exposed through food, drinking water, and dust, and pregnant women pass them across the placenta to their developing babies. While previous research had hinted that prenatal exposure to these metals could impair childhood neurodevelopment, the biological mechanisms remained murky.

To probe those mechanisms, researchers turned to the Wuhan Healthy Baby Cohort, a long-running birth cohort study in central China. The team, led by Ya Xie and Wei Xia of Wuhan Children’s Hospital and Huazhong University of Science and Technology, measured concentrations of thirteen rare earth elements in urine samples collected from mothers during the first trimester of pregnancy. They then performed metabolomic profiling on cord blood, the blood remaining in the placenta after birth, capturing a snapshot of each newborn’s metabolic state. Two years later, trained assessors evaluated each child’s cognitive and motor skills using the Bayley Scales of Infant Development, a widely used standardized test.

The results were striking. After adjusting for a range of covariates, the researchers found that higher first-trimester maternal levels of seven rare earth elements were each significantly associated with increased odds of cognitive developmental delay. The elements neodymium, samarium, gadolinium, holmium, erbium, thulium, and ytterbium each carried elevated risk, with adjusted odds ratios ranging from 1.13 to 1.25 for every one-unit increase in natural log-transformed urinary concentration. In practical terms, a modest rise in maternal exposure translated into a measurable increase in the likelihood that a toddler would score below developmental benchmarks on cognitive testing.

Because people are exposed to mixtures of rare earth elements rather than single metals in isolation, the team also used two statistical approaches designed for environmental mixtures: quantile g-computation and weighted quantile sum regression. Both methods confirmed a positive overall mixture effect, meaning the combined burden of rare earth elements raised the risk of both cognitive and psychomotor developmental delays. Intriguingly, the two methods identified different dominant contributors. Ytterbium emerged as the primary driver of cognitive developmental delay, while cerium played the leading role in psychomotor delay, hinting that individual rare earth elements may exert distinct biological effects even within a chemically similar family.

The study’s most innovative move came next. Rather than treating metabolites as isolated signals, the researchers applied weighted gene co-expression network analysis, a technique originally developed for genomics, to the cord blood metabolomic data. Using a soft-thresholding power of 5, which produced a network with strong scale-free topology, the team clustered thousands of metabolites into thirteen color-coded modules representing coordinated metabolic networks. This approach allowed them to ask which entire biological pathways, rather than individual molecules, were perturbed by prenatal rare earth exposure.

Two modules stood out. Prenatal rare earth element exposure was consistently associated with perturbations in metabolic modules enriched in lysophospholipids and phosphatidylethanolamines, two classes of phospholipids that are fundamental building blocks of cell membranes. These molecules are especially important in the developing brain, where membranes are being rapidly assembled and remodeled as neurons grow and form connections. The finding suggests that rare earth elements may interfere with fetal neurodevelopment by scrambling the lipid architecture on which brain construction depends.

Crucially, the association was more than correlational. Mediation analyses demonstrated that the lysophospholipid and phosphatidylethanolamine modules partially mediated the link between prenatal exposure and increased risk of neurodevelopmental delay. In other words, part of the statistical pathway from rare earth exposure to developmental impairment runs through these disrupted lipid networks. The identification of a plausible biological mechanism elevates the findings from an observational association toward a more causal interpretation, though the researchers note that mediation captures only a portion of the overall effect and other pathways likely remain to be discovered.

The implications ripple outward in several directions. For public health officials, the study strengthens the case for including rare earth elements in routine risk assessments during pregnancy, a step that has lagged far behind their growing environmental presence. As global demand for these metals surges with the green energy transition and the proliferation of electronic devices, exposure levels in ordinary populations may continue to climb. The study also points toward potential interventions: if phospholipid disruption is a key mechanism, researchers could eventually explore whether nutritional or pharmacological strategies that support healthy lipid metabolism during pregnancy might buffer developing brains against environmental insult.

For the families in Wuhan whose participation made the study possible, the research offers no easy answers, since rare earth exposure is diffuse and difficult for individuals to avoid. But it adds an urgent data point to a larger conversation about the hidden costs of the technologies that define modern life. The metals that power our digital world are now demonstrably reaching the next generation before they take their first breath, and the developing brain appears to register their presence in ways we are only beginning to measure. As the authors conclude, the findings highlight the need for health risk assessment of rare earth element exposure during pregnancy, and they provide early clues to intervention targets that could protect children before the damage is done.

Subject of Research: Prenatal rare earth element exposure and its association with neurodevelopmental delay and phospholipid metabolic disruption in children

Article Title: Cord blood metabolic network profiles link prenatal rare earth element exposure to neurodevelopmental delay in 2-year-old children

Article References: Xie, Y., Qiu, L., Fang, X., Cao, Z., Li, S., Gong, H., Tan, T., Wang, J., Li, Y., Xu, S., Xiao, H., Zhou, A., & Xia, W. (2026). Cord blood metabolic network profiles link prenatal rare earth element exposure to neurodevelopmental delay in 2-year-old children. BMC Medicine. https://doi.org/10.1186/s12916-026-05306-9

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05306-9

Keywords: rare earth elements, prenatal exposure, neurodevelopmental delay, metabolomics, cord blood, phospholipid metabolism, WGCNA, Bayley Scales, birth cohort, cognitive development, psychomotor development, environmental health

The tool, called the RNA Modular Expression Laboratory, or RNAmel, is described in BMC Bioinformatics by Karim Matmat and colleagues from the Department of Biomedicine at the University of Basel. On the surface, RNAmel belongs to a familiar genre: an interactive web application, built on R Shiny, that lets biologists click their way from a count matrix and a sample metadata table to publication-ready figures. But the authors argue that what sets their platform apart is not the interface itself but the architecture underneath it, and that architectural distinction speaks directly to one of the most persistent criticisms of graphical analysis tools in bioinformatics.

Most interactive analysis applications, the researchers note, are built as monolithic Shiny programs in which the analytical logic is entangled with the interface code. That coupling makes the analytical core hard to test, hard to script, and hard to reuse. It also obscures provenance: a scientist exploring results in a web browser may struggle to reconstruct exactly what transformations produced each number and each plot. Bespoke scripts avoid some of these problems but are notoriously difficult to reuse and audit across labs, especially when the original author has moved on. RNAmel takes a different route: it is implemented first as a tested R package, and the Shiny interface is deliberately kept as a thin wrapper over that package’s public application programming interface.

The consequences of that design are tangible. The analysis and figure-generation logic lives in a public function layer covered by more than 440 unit tests, meaning that the same tested functions drive the interactive session, the exported R script, and the automated test suite. In other words, the clickable interface and the command-line pipeline do not merely produce similar results; they are literally the same code. A user who performs a differential expression analysis through the browser can later export a runnable R script that reproduces every step, and a developer who modifies the package can rely on the test suite to catch regressions before they reach users. This is a rare level of formal verification for tools in this space, where unit testing is often an afterthought at best.

In terms of scope, RNAmel covers the full downstream bulk RNA-seq workflow on a single loaded dataset. Differential expression is handled by the widely used DESeq2 framework, with support for comparing multiple contrasts side by side. Exploratory dimensionality reduction is provided through principal component analysis and UMAP, using variance-stabilizing or regularized-logarithm transformations of the counts. Functional interpretation comes in several complementary flavors: gene-set enrichment analysis and over-representation analysis against resources such as Gene Ontology, KEGG, and the Molecular Signatures Database, along with per-sample signatures computed via GSVA and single-sample GSEA. The platform also integrates weighted gene co-expression network analysis through WGCNA, and infers transcription-factor and pathway activity using decoupleR. A notable and increasingly timely addition is an optional, opt-in layer of AI-assisted interpretation powered by large language models, which users can choose to enable rather than having it operate silently in the background.

Reproducibility, the paper’s central theme, is addressed on several fronts simultaneously. Analyses are recoverable as portable sessions, so a collaborator can pick up an analysis exactly where it left off. Exports come in three forms: a runnable R script, a Methods section of text that can be pasted directly into a manuscript, and a self-contained HTML report bundling the results. The entire computational environment is pinned down through a versioned Docker container and an accompanying renv.lock file that records the exact package versions, so that the same analysis run two years later, or on another institution’s server, will encounter the same software landscape. These features target the increasingly common scenario in which reviewers or journals ask authors to reproduce a figure that depends on a software stack nobody documented at the time.

To demonstrate that the platform actually performs as advertised, the authors put RNAmel through a battery of validation checks on two published human datasets. The first was a glucocorticoid-treatment study, a classic endocrinology experiment in which cells are perturbed by steroid signaling. The second was a larger and more heterogeneous benchmark: a 120-sample subset of The Cancer Genome Atlas spanning eight different cancer types, a scenario that stresses a tool’s ability to handle diverse sample compositions and broad biological variation. On these datasets, the team reported several layers of consistency checking, including exact reproduction of exported analyses, agreement with independent reference implementations of DESeq2 and WGCNA, and concordance with the limma-voom differential expression pipeline, an alternative method commonly used in the field.

That cross-method concordance deserves emphasis, because it addresses a question that haunts every multi-tool platform: when a single application bundles several statistical approaches, do the results actually match what a user would obtain by running those methods manually, with full attention and expert tuning? By showing that RNAmel’s outputs agree with independent implementations of the same algorithms, the authors provide evidence that the convenience of the integrated interface does not come at the cost of analytical fidelity. For a field in which differential expression results can vary noticeably across pipelines, this kind of explicit benchmarking is exactly the standard that integrated platforms should be held to.

The broader significance of RNAmel may lie in the precedent it sets for scientific software design. The argument implicit in the paper is that interactive accessibility and script-level transparency need not be a trade-off. Too often, tools that democratize analysis for bench scientists are dismissed as black boxes, while tools prized for rigor remain locked behind a command-line wall. By placing a tested, scriptable function layer beneath a friendly interface, RNAmel suggests a template for other domains of computational biology, from single-cell genomics to proteomics, where the same tension between usability and auditability plays out. The unit-test coverage of the analytical core, in particular, is a practice that many much larger bioinformatics projects have yet to adopt.

For the working biologist, the practical upshot is straightforward. RNAmel is free and open source, released under the MIT licence, and available through the GitHub repository of the first author. It runs in a versioned Docker environment, which shields users from the package-dependency conflicts that routinely derail bioinformatics workflows. Whether one chooses to interact with it through the browser or entirely through its R functions, the underlying analysis is the same tested code, and every step can be exported, documented, and replayed. In a discipline where reproducibility failures remain all too common, a tool that treats provenance as an architectural principle rather than an optional extra is a welcome arrival for anyone trying to turn sequencing data into trustworthy biology.

Subject of Research: An open-source R/Shiny platform for reproducible downstream bulk RNA-seq analysis and biological interpretation

Article Title: RNA modular expression laboratory (RNAmel): an integrated R/Shiny platform for reproducible bulk RNA-seq analysis and interpretation

Article References: Matmat, K., Pfanner, T., Silveira, R. P., & Wymann, M. P. (2026). RNA modular expression laboratory (RNAmel): an integrated R/Shiny platform for reproducible bulk RNA-seq analysis and interpretation. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06710-x

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06710-x

Keywords: RNA-seq, RNAmel, R Shiny, reproducibility, differential expression, DESeq2, WGCNA, gene set enrichment, decoupleR, transcriptomics, bioinformatics, open-source software

The study, published in BMC Plant Biology, focused on Bacillus safensis strain H23, a member of the plant growth-promoting rhizobacteria, a loose federation of root-colonizing microbes known to boost plant health. Led by Aliya Ayaz and corresponding author Huizhen Fu of Hainan University, with collaborators at Guangxi University, the team set out to answer a question that has long frustrated plant pathologists: what, at the molecular level, does a protective bacterium actually do inside a plant under attack? Biological control agents like Bacillus are widely used, but the transcriptional choreography they induce in the host has remained largely opaque, particularly for bacterial wilt, whose molecular basis of biocontrol the authors describe as poorly understood.

The experimental design was straightforward in concept but demanding in execution. The researchers grew tomato plants under different conditions: uninfected controls, plants infected with Ralstonia pseudosolanacearum, plants treated with the H23 bacterium, and plants that received both the protector and the pathogen. They then measured disease symptoms and a battery of physiological indicators, and harvested tissue at two time points, five and ten days into the interaction, to capture both the early and later phases of the plant’s response. RNA sequencing of these samples allowed the team to track how thousands of genes changed their activity across every treatment comparison.

The physiological results were unambiguous. Plants pretreated with H23 showed significantly reduced disease severity and better growth under pathogen pressure. The protective bacterium also enhanced the activity of the plant’s antioxidant enzymes, the first line of defense against the reactive oxygen molecules that both pathogens and stressed plant cells generate. Critically, H23 alleviated oxidative damage in measurable ways: levels of malondialdehyde, a marker of lipid damage in cell membranes, dropped, as did the accumulation of hydrogen peroxide and the leakage of electrolytes, which signals that membranes have lost their integrity. In short, the bacterium helped the plant keep its cellular machinery intact while the pathogen raged.

The transcriptomic data revealed the sheer scale of the underlying reprogramming. Across all treatment comparisons, the researchers identified 11,788 differentially expressed genes at the five-day time point and 11,674 at ten days, numbers indicating that the plant-microbe interaction triggers an extensive and sustained overhaul of the tomato genome’s activity rather than a brief defensive flare. Gene ontology and KEGG pathway enrichment analyses then showed that the two microbes pushed the plant in strikingly different directions. Treatment with H23 primarily induced genes involved in cell wall organization, polysaccharide metabolism, phenylpropanoid biosynthesis, MAPK signaling, and plant hormone signaling pathways. Infection with Ralstonia, by contrast, mainly activated oxidative stress responses, glutathione metabolism, and defense-related signaling.

That contrast is scientifically telling. The H23 profile suggests a bacterium that prepares the plant structurally and systemically: reinforcing cell walls, ramping up phenylpropanoid biosynthesis, which produces lignin and other defensive compounds, and priming hormone signaling networks that coordinate immune responses. The pathogen profile, meanwhile, reads like the signature of a plant under oxidative siege, scrambling to neutralize reactive oxygen species through glutathione metabolism and to mount whatever defense signaling it can. The implication is that H23 does not simply fight the pathogen directly; it shifts the plant into a state of induced systemic resistance, a primed defensive posture in which the plant’s own architecture and signaling systems are fortified before and during attack.

To make sense of more than eleven thousand changing genes, the team turned to Weighted Gene Co-expression Network Analysis, or WGCNA, a statistical framework that groups genes into modules based on how their expression patterns correlate across samples. The analysis identified five key modules whose activity was associated with the physiological traits the researchers had measured. Within these modules, the team pinpointed 150 hub genes, the most highly connected and presumably most influential members of each co-expression cluster. These hub genes were associated with disease resistance, oxidative stress regulation, detoxification, hormone signaling, and secondary metabolism, painting a coherent picture of the biological processes that matter most in the protected plants.

The final and perhaps most novel step was the application of machine learning. Rather than relying on enrichment statistics alone, the researchers fed their candidate gene data into machine-learning algorithms to prioritize the genes most strongly associated with pathogen response and with H23-mediated regulation. The analysis converged on ten key candidate genes, a compact set that the authors propose as core molecular markers of the protective interaction. This combination of network analysis and machine learning reflects a broader trend in plant biology: as sequencing datasets balloon, computational triage has become essential for distinguishing the handful of genes that truly drive a phenotype from the thousands that merely change along the way.

The practical implications extend in two directions. For breeders, the ten candidate genes offer concrete targets: markers that could be tracked in breeding programs aimed at tomato varieties with stronger innate resistance to bacterial wilt, whether through conventional selection or genome editing. For agronomists and integrated pest management specialists, the study strengthens the case for PGPR-based biocontrol as a component of integrated disease management, providing a molecular rationale for what has often been an empirically observed benefit. Because the protective effect operates through the plant’s own antioxidant defenses, structural reinforcement, and immune signaling, it complements rather than duplicates existing control measures, and it may reduce dependence on chemical treatments that struggle against a root-dwelling, vascular pathogen.

The research, supported by funding from Hainan Province and Hainan University, also carries the usual caveats of early-stage translational science. The candidate genes were identified through association and prioritization, and confirming their function will require direct genetic validation in tomato. Field performance of H23 across diverse soils, climates, and pathogen strains remains to be demonstrated. Yet the study’s framework, pairing physiological measurement with time-resolved transcriptomics, co-expression networks, and machine learning, offers a template that can be applied to other beneficial microbes and other crops. For a disease as destructive and intractable as bacterial wilt, a molecular map of how a harmless soil bacterium teaches a tomato plant to defend itself is a meaningful step toward crops that can stand firm where they once collapsed.

Subject of Research: Bacillus safensis-mediated induction of tomato resistance to Ralstonia bacterial wilt revealed by transcriptomics and machine learning

Article Title: Integrated transcriptomics and machine learning reveal core genes associated with Bacillus safensis-mediated resistance to tomato bacterial wilt

Article References: Ayaz, A., Ikram, M., Jin, R., Gao, F., Yu, W., Wang, Z., Liu, P., Huang, J., & Fu, H. (2026). Integrated transcriptomics and machine learning reveal core genes associated with Bacillus safensis-mediated resistance to tomato bacterial wilt. BMC Plant Biology. https://doi.org/10.1186/s12870-026-10116-5

Image Credits: AI Generated

DOI: 10.1186/s12870-026-10116-5

Keywords: tomato, bacterial wilt, Bacillus safensis, Ralstonia pseudosolanacearum, PGPR, transcriptomics, differentially expressed genes, WGCNA, machine learning, plant immunity, biological control, antioxidant defense