GLIO-Select: Machine Learning-Based Feature Selection and Weighting of Tissue and Serum Proteomic and Metabolomic Data Uncovers Sex Differences in Glioblastoma

International Journal of Molecular Sciences, Journal Year: 2025, Volume and Issue: 26(9), P. 4339 - 4339

Published: May 2, 2025

Glioblastoma (GBM) is a fatal brain cancer known for its rapid and aggressive growth, with some studies indicating that females may have better survival outcomes compared to males. While sex differences in GBM been observed, the underlying biological mechanisms remain poorly understood. Feature selection can lead identification of discriminative key biomarkers by reducing dimensionality from high-dimensional medical datasets improve machine learning model performance, explainability, interpretability. uncover unique sex-specific biomarkers, determinants, molecular profiles patients GBM. We analyzed proteomic metabolomic serum biospecimens obtained 109 pathology-proven glioblastoma on NIH IRB-approved protocols full clinical annotation (local dataset). Serum analysis was performed using Somalogic aptamer-based technology (measuring 7289 proteins) metabolome University Florida’s SECIM (Southeast Center Integrated Metabolomics) platform 6015 metabolites). Machine learning-based feature employed identify proteins metabolites associated male female labels datasets. Results were publicly available (CPTAC TCGA) same methodology TCGA data previously structured glioma grading. Employing hybrid approach, utilizing both LASSO mRMR, conjunction rank-based weighting method (i.e., GLIO-Select), we linked purposes reduction used separate set explore possible linkages between mutations tumor identified several hundred features male/female class label Using local serum-based dataset patients, 17 (100% ACC) 16 (92% datasets, respectively. CPTAC tissue-based (8828 59 features), 5 (99% 13 (80% The or tissue (CPTAC) achieved highest accuracy rates 99%, respectively), followed metabolome. yielded clinically (PSA, PZP, HCG, FSH) which distinct (RPS4Y1 DDX3Y), providing methodological validation, PZP defensins (DEFA3 DEFB4A) representing shared tissue. Metabolomic homocysteine pantothenic acid. Several signals emerged are be but not sex, requiring further research, as well novel either glioma. EGFR, FAT4, BCOR three 64% ACC grading set. GLIO-Select shows remarkable results when different types (e.g., tissue-based) our analyses. proposed approach successfully reduced relevant less than twenty each dataset. appear highly effective identifying biologically These findings suggest noninvasive biospecimen-based analyses provide more accurate detailed insights into variable (SABV) other biospecimens, linking pathology via immune response, amino acid metabolism, hallmark research. Our underscore importance biospecimen choice enhancing interpretation omics understanding sex-based This discovery holds significant potential personalized treatment plans patient outcomes.

Language: Английский

---

Radiomics and AI-Based Prediction of MGMT Methylation Status in Glioblastoma Using Multiparametric MRI: A Hybrid Feature Weighting Approach

Diagnostics, Journal Year: 2025, Volume and Issue: 15(10), P. 1292 - 1292

Published: May 21, 2025

Background/Objectives: Glioblastoma (GBM) is a highly aggressive primary central nervous system tumor with median survival of 14 months. MGMT (O6-methylguanine-DNA methyltransferase) promoter methylation status key biomarker as prognostic indicator and predictor chemotherapy response in GBM. Patients methylated disease progress later survive longer (median rate 22 vs. 15 months, respectively) compared to patients unmethylated disease. GBM undergo an MRI the brain prior diagnosis following surgical resection for radiation therapy planning ongoing follow-up. There currently no imaging Studies have attempted connect appearance determine if can be leveraged provide information non-invasively more expeditiously. Methods: Artificial intelligence (AI) identify features that are not distinguishable human eye linked status. We employed UPenn-GBM dataset whom was available (n = 146), employing novel radiomic method grounded hybrid feature selection weighting predict Results: The best classification result obtained resulted mean accuracy value 81.6% utilizing 101 selected five-fold cross-validation. Conclusions: This favorably similar studies literature. Validation external datasets remains critical enhance generalizability propagate robust results while reducing bias. Future directions include multi-channel data integration deep ensemble learning methods improve predictive performance.

Language: Английский

---

MetaWise: Combined Feature Selection and Weighting Method to Link the Serum Metabolome to Treatment Response and Survival in Glioblastoma

International Journal of Molecular Sciences, Journal Year: 2024, Volume and Issue: 25(20), P. 10965 - 10965

Published: Oct. 11, 2024

Glioblastoma (GBM) is a highly malignant and devastating brain cancer characterized by its ability to rapidly aggressively grow, infiltrating tissue, with nearly universal recurrence after the standard of care (SOC), which comprises maximal safe resection followed chemoirradiation (CRT). The metabolic triggers leading reprogramming tumor behavior resistance are an area increasingly studied in relation molecular features associated outcome. There currently no metabolomic biomarkers for GBM. Studying alterations GBM patients undergoing CRT could uncover biochemical pathways involved response resistance, identification novel optimization treatment response. feature selection process identifies key factors improve model’s accuracy interpretability. This study utilizes combined approach, incorporating both Least Absolute Shrinkage Selection Operator (LASSO) Minimum Redundancy–Maximum Relevance (mRMR), alongside rank-based weighting method (i.e., MetaWise) link 12-month 20-month overall survival (OS) status Our shows promising results, reducing dimensionality when employed on serum-based large-scale datasets (University Florida) all our analyses. proposed successfully identified set eleven serum shared among three datasets. computational results show that utilized achieves 96.711%, 92.093%, 86.910% rates 48, 46, 33 selected CRT, 12-month, OS-based datasets, respectively. discovery has implications developing personalized plans improving patient outcomes.

Language: Английский

---

Serum CD133-Associated Proteins Identified by Machine Learning Are Connected to Neural Development, Cancer Pathways, and 12-Month Survival in Glioblastoma

Cancers, Journal Year: 2024, Volume and Issue: 16(15), P. 2740 - 2740

Published: Aug. 1, 2024

Glioma is the most prevalent type of primary central nervous system cancer, while glioblastoma (GBM) its aggressive variant, with a median survival only 15 months when treated maximal surgical resection followed by chemoradiation therapy (CRT). CD133 potentially significant GBM biomarker. However, current clinical biomarker studies rely on invasive tissue samples. These make prolonged data acquisition impossible, resulting in increased interest use liquid biopsies. Our study, analyzed 7289 serum proteins from 109 patients pathology-proven obtained prior to CRT using aptamer-based SOMAScan® proteomic assay technology. We developed novel methodology that identified 24 linked both and 12-month overall (OS) through multi-step machine learning (ML) analysis. were subsequently subjected clustering evaluations, categorizing into five risk groups accurately predicted OS based their protein profiles. Most these are involved brain function, neural development, and/or cancer biology signaling, highlighting significance potential predictive value. Identifying provides valuable foundation for future investigations as validation clinically applicable biomarkers can unlock immense diagnostics treatment monitoring.

Language: Английский

---