miRNA liquid biopsy combined with MRI radiomics for improved outcome prediction in glioblastoma: integrated machine learning analysis of longitudinal data from 73 patients

Neuro-Oncology Advances · 2026-05-14

Abstract Background While both MRI radiomics and miRNA liquid biopsy have shown promise for glioblastoma prognostication, state-of-the-art methods may enable novel integration of multimodal data to further optimize performance. Methods Serum samples (n = 193) were collected from 73 patients with pathology-confirmed glioblastoma at a single center. We quantified 798 miRNAs per sample using nCounter. Radiomic features were automatically extracted from MRI scans (n = 306) obtained during the same follow-up period. A new data integration pipeline was applied to evaluate machine learning-based outcome predictions using miRNA, radiomic, and miRNA+radiomic input datasets. All models included a set of clinical covariates of known prognostic value. Time-averaged AUC analysis was used to incorporate longitudinal sampling. In experiments to classify postoperative samples labeled by likely disease burden, performance-driving miRNAs were further investigated using counterfactual analysis (CA) and Shapley Additive Explanations (SHAP). Results The recurrence rate was 75% (median 222 days, IQR [93–378]), and post-recurrence mortality was 62% (412 days, [274–695]) during the follow-up period. Radiomics outperformed miRNAs for recurrence prediction (time-averaged AUC = 0.66 [0.60–0.71] versus AUC = 0.56 [0.53–0.59]), while miRNAs outperformed radiomics for survival prediction (AUC = 0.70 [0.64–0.77] versus AUC = 0.55 [0.47–0.60]). Combined miRNA+radiomics models performed well for recurrence (AUC = 0.64 [0.44–0.80]) and best overall for survival (AUC = 0.76 [0.63–0.86]). Several performance-driving miRNAs were also identified on CA and SHAP analyses. Conclusions Serum miRNA profiling combined with MRI radiomics may improve longitudinal approximation of postoperative glioblastoma prognosis. Integrating multimodal data is feasible and could enable more informed counseling of patients and families over the disease course.

Transcriptomic Reprogramming from Neural Stem Cells to Glioblastoma Stem Cells Reveals Partially Independent Roles of Alternative Splicing and Transposable Elements

Brown Digital Repository · 2026-05-05

A long-read RNA sequencing and polysome profiling framework reveals transposable element–driven transcript diversity and translational rewiring in glioblastoma

bioRxiv (Cold Spring Harbor Laboratory) · 2026-04-21

Abstract Background Transposable elements (TEs) account for over half of the human genome and are often derepressed in cancer. TEs can add cryptic splice sites, undergo exonization, and generate gene–TE fusion transcripts, but the combined effects of TEs on RNA processing and translation in glioblastoma stem cells (GSCs) remains incompletely elucidated. Results We combined long-read RNA sequencing with polysome profiling in four patient-derived GSCs and two neural stem cell (NSC) controls to resolve TE-associated transcript diversity and its relationship to ribosomal engagement. Across GSCs, we identified 13,421 alternative splicing (AS) events, 3,077 of which contained TEs within 150 bp of splice junctions. AS sites proximal to TEs were associated with increased isoform switching compared to non–TE-associated AS sites (odds ratio 2.9 - 4.3). Moreover, AS isoforms generated from TE-proximal sites were more likely to exhibit altered ribosomal association (odds ratio 2.54). Directional shifts were observed, with shorter isoforms associating with monosome fractions and longer isoforms with polysome fractions. To enable systematic detection of gene - TE chimeric transcripts, we developed FuTER (Fusion TE Reporter), a long-read–based framework for identifying TE-associated fusions. Application to GSC datasets identified 78 GSC enriched fusion transcripts, several supported by breakpoint-spanning reads in polysome fractions, consistent with ribosome association. Conclusions Our data suggest that TEs correlate with abnormal splicing activity and altered ribosome engagement in glioblastoma stem cells. By integrating long-read sequencing with polysome profiling and fusion detection, we establish a framework for analysis of TE-induced transcript diversity and its effects on cancer evolution and plasticity.

103 MiRNA Profiling on Liquid Biopsy Outperforms Automated MRI Radiomics for Patient Outcome Prediction in Glioblastoma: Integrated Machine Learning Analysis of Longitudinal, Multimodal Data From 73 Patients

Neurosurgery · 2026-03-26

Multiomic profiling of hypoxic glioblastoma stem cells reveals expansion of subpopulations with distinct epigenetic and CNV profiles

bioRxiv (Cold Spring Harbor Laboratory) · 2025-05-09 · 1 citations

Abstract Glioblastoma is characterized by extensive intratumoral heterogeneity driven by dynamic mi-croenvironmental cues such as hypoxia. While transcriptional and epigenetic variability have been separately linked to hypoxia responses, the integrated impact of hypoxia on gene regulation and clonal architecture in glioblastoma stem cells (GSCs) remains poorly defined. We applied singlenucleus multi-omics—integrating RNA-seq and ATAC-seq—to patient-derived GSCs cultured under normoxic or hypoxic conditions. This enabled simultaneous profiling of gene expression and chromatin accessibility within the same cells. Transcription factor (TF) regulatory networks were inferred using Dictys, while RNA-chromatin dynamics were modeled with MultiVelo. Clonal structure and copy number variations (CNVs) were resolved at single-cell resolution using RIDDLER on snATAC-seq data. Hypoxia induced the emergence of four distinct GSC subpopulations with unique transcriptomic and epigenetic profiles enriched for mesenchymal, angiogenic, and proliferative signatures. Regulatory network modeling revealed novel hypoxia-associated TFs—SP2, CREM, and ETV3—that modulate downstream oncogenic pathways. Trajectory analysis uncovered hypoxia-driven reversals in RNA-chromatin coupling, revealing dysregulated future transcriptional states of key genes such as MMP16 and SVIL. CNV profiling identified 13 clonal substructures, with specific clones (e.g., 5, 6, 9) selectively enriched under hypoxia and harboring distinct chromosomal alterations. These results demonstrate coordinated remodeling of GSC gene regulation and clonal fitness in response to hypoxic stress. Our findings reveal that hypoxia drives concurrent epigenetic, transcriptomic, and clonal selection in glioblastoma stem cells. This integrated model of hypoxia-induced plasticity provides mechanistic insights into tumor adaptation and identifies novel regulators that may serve as targets for therapeutic intervention in the hypoxic niche of glioblastoma.

Machine learning on multiple epigenetic features reveals H3K27Ac as a driver of gene expression prediction across patients with glioblastoma

PLoS Computational Biology · 2025-08-07 · 2 citations

Epigenetic mechanisms play a crucial role in driving transcript expression and shaping the phenotypic plasticity of glioblastoma stem cells (GSCs), contributing to tumor heterogeneity and therapeutic resistance. These mechanisms dynamically regulate the expression of key oncogenic and stemness-associated genes, enabling GSCs to adapt to environmental cues and evade targeted therapies. Importantly, epigenetic reprogramming allows GSCs to transition between cellular states, including therapy-resistant mesenchymal-like phenotypes, underscoring the need for epigenetic-targeting strategies to disrupt these adaptive processes. Understanding these epigenetic drivers of gene expression provides a foundation for novel therapeutic interventions aimed at eradicating GSCs and improving glioblastoma outcomes. Using machine learning (ML), we employ cross-patient prediction of transcript expression in GSCs by combining epigenetic features from various sources, including ATAC-seq, CTCF ChIP-seq, RNAPII ChIP-seq, H3K27Ac ChIP-seq, and RNA-seq. We investigate different ML and deep learning (DL) models for this task and ultimately build our final pipeline using XGBoost. The model trained on one patient generalizes to other 11 patients with high performance. Notably, H3K27Ac alone from a single patient is sufficient to predict gene expression in all 11 patients. Furthermore, the distribution of H3K27Ac peaks across the genomes of all patients is remarkably similar. These findings suggest that GSCs share a common distributional pattern of enhancer activity characterized by H3K27Ac, which can be utilized to predict gene expression in GSCs across patients. In summary, while GSCs are known for their transcriptomic and phenotypic heterogeneity, we propose that they share a common epigenetic pattern of enhancer activation that defines their underlying transcriptomic expression pattern. This pattern can predict gene expression across patient samples, providing valuable insights into the biology of GSCs.

BIOM-67. A MIRNA-BASED LIQUID BIOPSY FOR BRAIN TUMOR DIAGNOSIS AND IN VITRO EFFECTS OF MIRNA INHIBITION ON GLIOBLASTOMA STEM CELL GROWTH, VIABILITY, AND SELF-RENEWAL

Neuro-Oncology · 2025-11-01

Abstract INTRODUCTION Blood biomarkers could enable cost-effective, early detection of brain tumors. While circulating miRNAs are promising candidates, disease specificity is poorly understood. Here, we use an unbiased approach to identify miRNAs which distinguish brain tumor cases from controls, correlate findings with glioblastoma stem cell (GSC) miRNA expression, and characterize the effects of inhibiting select miRNAs on GSC phenotypes. METHODS Patients undergoing surgery at a single center for newly diagnosed brain tumors (pathology-confirmed glioblastoma n=21, meningioma n=27, or brain metastasis n=9) and age-matched controls undergoing elective surgery for degenerative spinal disease (DSD n=24, e.g. microdiscectomy) were recruited for preoperative research blood collection (2019–2024). 798 miRNAs were quantified per serum sample using nCounter (Nanostring). AutoML (JADBio) was applied to identify miRNAs which optimally classify samples by these diagnoses, and model stability analyses confirmed those that classify samples most consistently. Model performance is reported as area under receiver operating characteristic (AUROC) and mean precision (MP). The same miRNAs were quantified in GSC media. Those which distinguished glioblastoma cases from controls were inhibited in vitro via lipid transfection (qPCR-confirmed) to observe effects on GSC growth (Incucyte), viability (CellTiter-Glo®), and self-renewal (sphere formation). RESULTS miRNA-based classification of brain tumor (n=57) versus DSD controls (n=24) yielded excellent performance (AUROC=0.93 [0.88–0.98], MP=0.95 [0.91–0.98]), as did glioblastoma versus DSD (AUROC=0.94 [0.86–1.00], MP=0.95 [0.91–1.00]) and glioblastoma/metastasis versus DSD (AUROC=0.92 [0.85–0.97], MP=0.95 [0.91–0.98]). Some performance-driving miRNAs (miR-150-5p, miR-1283) were under-expressed in glioblastoma cases while others (miR-873-3p, let-7b-5p) were over-expressed. These miRNAs were also expressed in GSC media. Inhibiting let-7b-5p was noted to reduce GSC growth, viability, and self-renewal, while inhibiting others augmented these phenotypes. CONCLUSIONS A preoperative miRNA signature may accurately distinguish brain tumor patients from age-matched controls. Inhibition of these miRNAs affected GSC phenotypes. Future studies will validate classification models in an independent cohort and investigate the effect of miRNA inhibition on GSC transcript expression (RNA-sequencing).

DDDR-11. Pharmacological inhibition of or reduced EZH2 levels sensitized diffuse intrinsic pontine gliomas (DIPG) to ONC201, leading to synthetic lethality

Neuro-Oncology · 2025-11-01

Abstract BACKGROUND ONC201 is a first-in-class, blood-brain barrier penetrant imipridone that showed promise against H3K27M gliomas. Mechanistically, H3K27M mutation inhibits EZH2 in the methylating H3K27. In this context, we hypothesized that pharmacologic inhibition of EZH2 should recapitulate the physiologic effects of H3K27M mutation, with the corollary of synthetic lethal interactions between EZH2 inhibition and ONC201. OBJECTIVE Here, we investigated i) EZH2 expression level in H3K27 wild-type diffuse intrinsic pontine gliomas correlated with ONC201 sensitivity and ii) potential synergy between ONC201 and EZH2 inhibition. METHODS pre-clinical laboratory investigation. RESULTS In a panel of H3K27 wild-type diffuse intrinsic pontine glioma cell lines, EZH1/EZH2 expression correlated with ONC201 sensitivity. RNA-seq showed that ONC201 and EHZ2 inhibitor tazemetostat-treated cells exhibited similar transcriptional profiles, sharing top-regulated genes. This finding suggests that ONC201 and EHZ2 inhibition converge on regular nodes within the same linear pathway to disrupt shared cellular functions. Supporting this hypothesis, ONC201 and EZH2i-treatment caused similar changes in the profile of cytokine release. In contrast, there was no overlap in the transcription or cytokine profiles obtained after ONC201 and Panobinostat (an HDAC inhibitor) treatment. Across all H3K27 wild-type diffuse intrinsic pontine glioma cell lines, the combination of ONC201 and the EZH2 inhibitor tazemetostat resulted in synergistic cytotoxic effects. Notably, the H3K27 methylation function of EZH2 was not affected by ONC201 treatment, indicating this function does not define the principal convergence point for the physiologic effects of ONC201 and EZH2 inhibition. CONCLUSION ONC201 and EHZ2 inhibition converge on nodes within the same linear pathway and exhibit synthetic lethal interactions. These findings bear therapeutic implications and provide the foundation for drug combinations with ONC201.

Supplementary Table 1 from Chi3l1 Is a Modulator of Glioma Stem Cell States and a Therapeutic Target in Glioblastoma

2025-11-24

<p>Supplementary Table 1</p>

Innovative Method for Fully Automated, Enzyme-Free Tissue Dissociation and Preparation for Single-Cell Analysis

Cellular and Molecular Bioengineering · 2025-07-03 · 3 citations