Information-Based Medicine in Glioma Patients: A Clinical Perspective

Imaging Biomarkers of Glioblastoma Treatment Response: A Systematic Review and Meta-Analysis of Recent Machine Learning Studies

OBJECTIVE

Monitoring biomarkers using machine learning (ML) may determine glioblastoma treatment response. We systematically reviewed quality and performance accuracy of recently published studies.

METHODS

Following Preferred Reporting Items for Systematic Reviews and Meta-Analysis: Diagnostic Test Accuracy, we extracted articles from MEDLINE, EMBASE and Cochrane Register between 09/2018-01/2021. Included study participants were adults with glioblastoma having undergone standard treatment (maximal resection, radiotherapy with concomitant and adjuvant temozolomide), and follow-up imaging to determine treatment response status (specifically, distinguishing progression/recurrence from progression/recurrence mimics, the target condition). Using Quality Assessment of Diagnostic Accuracy Studies Two/Checklist for Artificial Intelligence in Medical Imaging, we assessed bias risk and applicability concerns. We determined test set performance accuracy (sensitivity, specificity, precision, F1-score, balanced accuracy). We used a bivariate random-effect model to determine pooled sensitivity, specificity, area-under the receiver operator characteristic curve (ROC-AUC). Pooled measures of balanced accuracy, positive/negative likelihood ratios (PLR/NLR) and diagnostic odds ratio (DOR) were calculated. PROSPERO registered (CRD42021261965).

RESULTS

Eighteen studies were included (1335/384 patients for training/testing respectively). Small patient numbers, high bias risk, applicability concerns (particularly confounding in reference standard and patient selection) and low level of evidence, allow limited conclusions from studies. Ten studies (10/18, 56%) included in meta-analysis gave 0.769 (0.649-0.858) sensitivity [pooled (95% CI)]; 0.648 (0.749-0.532) specificity; 0.706 (0.623-0.779) balanced accuracy; 2.220 (1.560-3.140) PLR; 0.366 (0.213-0.572) NLR; 6.670 (2.800-13.500) DOR; 0.765 ROC-AUC.

CONCLUSION

ML models using MRI features to distinguish between progression and mimics appear to demonstrate good diagnostic performance. However, study quality and design require improvement.

Rosmarinic acid inhibits cell proliferation, migration, and invasion and induces apoptosis in human glioma cells

There is a growing evidence that Fyn kinase is upregulated in glioblastoma multiforme (GBM), where it plays a key role in tumor proliferation and invasion. In the present study, the antitumor effects of rosmarinic acid (RA), a Fyn inhibitor, were explored in human‑derived U251 and U343 glioma cell lines. These cells were treated with various concentrations of RA to determine its effects on proliferation, migration, invasion, apoptosis, and gene and protein expression levels. The CCK‑8 assay revealed that RA significantly suppressed cell viability of U251 and U343 cells. Furthermore, RA significantly reduced proliferation rates, inhibited migration and invasion, and decreased the expression levels of invasion‑related factors, such as matrix metalloproteinase (MMP)‑2 and MMP‑9. TUNEL staining revealed that RA resulted in a dose‑dependent increase of U251 and U343 cell apoptosis. In line with this finding, the expression of apoptosis suppressor protein Bcl‑2 was downregulated and that of the pro‑apoptotic proteins Bax and cleaved caspase‑3 was increased. In addition, it was revealed that the phosphatidylinositol 3‑kinase (PI3K)/Akt/nuclear factor‑κB (NF‑κB) signaling pathway was involved in RA‑induced cytotoxicity in U251 and U343 cells. Collectively, the present study suggested RA as a drug candidate for the treatment of GBM.

Identification and validation of potential novel prognostic biomarkers for patients with glioma based on a gene co-expression network

BACKGROUND

Glioma is widely regarded as one of most lethal and challenging diseases of the nervous system. The aim of this study was to identify novel biomarkers that offer better prognosis prediction for Chinese patients with glioma.

METHODS

By using systematic approaches, the co-expression modules were identified from the Chinese Glioma Genome Atlas (CGGA) database through weighted gene co-expression network analysis and functional enrichment of essential modules of Kyoto Encyclopedia of Genes and Genomes terms. The co-expression modules were validated using The Cancer Genome Atlas database and the protein-protein interaction (PPI) network.

RESULTS

For network construction, 5,374 among 21,494 genes were selected, and an increasing genetic variance was associated with the prognosis of glioma. By using functional enrichment analysis, the involvement of multiple vital processes, including metabolism of fatty acids, was correlated with the patient prognosis. Notably, five hub genes (KCNB1, UST, SOX8, KLHL42, and HDAC4) were identified for these processes. Accordingly, using the Kaplan-Meier method, there was enhanced expression of these genes in patients with significantly lower overall survival rates, especially those from the CGGA database.

CONCLUSIONS

This study not only revealed the essential co-expression gene modules in patients with glioma, but it also unraveled the potential signaling pathways underlying these functional processes.

Next Generation Sequencing and Machine Learning Technologies Are Painting the Epigenetic Portrait of Glioblastoma

Even with a rare occurrence of only 1.35% of cancer cases in the United States of America, brain tumors are considered as one of the most lethal malignancies. The most aggressive and invasive type of brain tumor, glioblastoma, accounts for 60–70% of all gliomas and presents with life expectancy of only 12–18 months. Despite trimodal treatment and advances in diagnostic and therapeutic methods, there are no significant changes in patient outcome. Our understanding of glioblastoma was significantly improved with the introduction of next generation sequencing technologies. This led to the identification of different genetic and molecular subtypes, which greatly improve glioblastoma diagnosis. Still, because of the poor life expectancy, novel diagnostic, and treatment methods are broadly explored. Epigenetic modifications like methylation and changes in histone acetylation are such examples. Recently, in addition to genetic and molecular characteristics, epigenetic profiling of glioblastomas is also used for sample classification. Further advancement of next generation sequencing technologies is expected to identify in detail the epigenetic signature of glioblastoma that can open up new therapeutic opportunities for glioblastoma patients. This should be complemented with the use of computational power i.e., machine and deep learning algorithms for objective diagnostics and design of individualized therapies. Using a combination of phenotypic, genotypic, and epigenetic parameters in glioblastoma diagnostics will bring us closer to precision medicine where therapies will be tailored to suit the genetic profile and epigenetic signature of the tumor, which will grant longer life expectancy and better quality of life. Still, a number of obstacles including potential bias, availability of data for minorities in heterogeneous populations, data protection, and validation and independent testing of the learning algorithms have to be overcome on the way.