Abstract Background: Angiogenesis, the formation of new blood vessels, is central to glioma progression and supports tumor growth and invasion. Early identification of aggressive biology in lower-grade gliomas (LGG) is essential for guiding treatment decisions. We developed an immune-oriented angiogenesis gene signature that characterizes aggressive tumor phenotypes in LGG and glioblastoma (GBM). The study also evaluates its association with key glioma biomarkers, including IDH mutations and 1p/19q co-deletions, and assesses its prognostic relevance across both tumor types. Objective: To develop and validate an immune-oriented angiogenesis gene signature that identifies aggressive glioma biology, predicts patient prognosis in LGG and GBM, and evaluates its relationship with IDH mutation and 1p/19q co-deletion status. Methods: We analyzed LGG and GBM samples from The Cancer Genome Atlas (TCGA). Immune infiltration levels were estimated using EPIC, and samples were classified into high and low infiltration groups. Differentially expressed genes between these groups were identified, followed by Weighted Gene Co-expression Network Analysis (WGCNA) to cluster co-expressed genes into modules. Fisher’s exact test was used to assess enrichment of WGCNA module genes within the Hallmark Angiogenesis gene set, prioritizing overlapping genes. Associations with IDH status, 1p/19q co-deletion, and patient survival were evaluated using Cox regression, Kaplan-Meier analysis, and time-dependent ROC curves. Validation was performed in the Chinese Glioma Genome Atlas (CGGA) and two single-cell datasets (pLGG: GSE222850; GBM: GSE138794). Results: The derived immune-oriented angiogenesis signature was significantly associated with IDH mutation and 1p/19q subtypes (Adj. p ≤ 0.05) and strongly predicted patient outcomes. Higher signature expression correlated with poorer survival and remained independently prognostic in multivariate analyses (e.g., TCGA LGG p = 3.28×10-7; CGGA LGG p = 1.08×10-8). The signature showed strong predictive accuracy in time-dependent ROC analyses (TCGA LGG AUCs: 1-year 0.959, 3-year 0.873, 5-year 0.812). Gene expression patterns observed in LGG showed a progressive trend consistent with certain GBM subgroups. In single-cell datasets, most genes showed significant differential expression between high and low immune infiltration groups (Adj. p ≤ 0.05). Conclusion: We identified and validated an immune-oriented angiogenesis gene signature that integrates genetic, molecular, and clinical features across gliomas. The signature demonstrates strong prognostic value in both LGG and GBM and reflects increasing angiogenic activity along glioma progression. This gene set may serve as a practical tool for patient risk stratification and offers potential targets for future therapeutic strategies. Citation Format: Pijush Das, Kevin A. Camphausen, Uma Shankavaram. Prognostic relevance of angiogenesis-associated genes in gliomas: Influence of idh/1p19q status and opportunities for antiangiogenic immunotherapy strategies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4794.

5110 Background: High-plex serum proteomics can track prostate cancer biology and treatment response beyond PSA, but interpretation is limited by high dimensionality. Pathway analysis can help, but incomplete overlap between legacy pathway libraries and assay panels can yield hard-to-interpret enrichments dominated by a few measured proteins. We evaluated an automated large language model (LLM) workflow using protein annotations to build protein sets containing only measured proteins. Methods: Serum from 88 individuals was profiled with an aptamer-based ~7,000-protein assay (SomaScan 7K; SomaLogic, USA). The cohort included localized prostate cancer treated with radiotherapy (RT) with or without androgen-deprivation therapy (ADT) with serial sampling (pre-RT n = 76, end-RT n = 72, ~1-month follow-up n = 76), plus metastatic (n = 4) and normal controls (n = 8). Assay fidelity was assessed by correlating PSA aptamers with clinical PSA. Protein-level analyses tested ADT effects and paired within-patient changes across RT. For program-level analysis, UniProt annotations for each measured protein were processed with an LLM to generate structured summaries, converted to embeddings, and clustered into protein sets restricted to SomaScan proteins. Set coherence and assay coverage were compared to Gene Ontology (GO) and Reactome mappings. Protein-set enrichment comparing ADT vs no ADT identified candidate programs and were evaluated for association with biochemical recurrence among high-risk ADT-treated patients (n = 50; 20 events) using Cox models adjusted for pre-treatment PSA. Results: PSA aptamers correlated with clinical PSA (Spearman r = 0.66–0.77). ADT suppressed reproductive-axis proteins (LH, FSH, hCG) and prostate-lineage proteins (PSA, PAP, TGM4); RT contrasts captured acute epithelial injury/lymphoid suppression followed by remodeling and stress responses. LLM protein sets showed higher set name/protein description coherence (median cosine similarity 0.59) than GO (0.30) or Reactome (0.36) and covered all measured proteins; GO/Reactome sets averaged ~50% member coverage. Pre-RT ADT enrichment identified histone programs (NES 2.13–2.18; FDR < 0.01), summarized as a Histone H2 score. Histone H2 score increased across no ADT, ADT, and metastatic samples (Kruskal–Wallis p = 0.003; all pairwise FDR < 0.05). In high-risk ADT-treated patients, pre-RT Histone H2 score was associated with recurrence (HR 0.53, FDR = 0.024), and remained significant in a bivariate model with pre-treatment PSA (Histone H2 HR 0.49, FDR = 0.013; PSA HR 2.78, FDR < 0.001). Non-H2 histone score (H1/H3) showed no ADT-associated shift (Wilcoxon p = 0.70), supporting specificity of the Histone H2 signal. Conclusions: LLM-derived protein sets restricted to measured proteins improved interpretability of serum proteomics and revealed an ADT-associated Histone H2 program complementary to PSA.

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

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

Background: Accurate survival prediction in cancer patients is crucial for personalized oncology, directly influencing treatment decisions and patient care. Integrating multi-omics data with clinical information can significantly enhance prognostic models. However, existing methods often necessitate extensive manual preprocessing and specialized bioinformatics expertise, limiting their practical use in research and clinical settings. This study presents the Cancer Patient Survival Model (CPSM), an R package providing an automated, streamlined pipeline for generating reliable, interpretable survival predictions from complex, high-dimensional data. Objective: The objective of this case study is to evaluate the utility of CPSM in analyzing survival probabilities and identifying prognostic markers for Kidney Renal Clear Cell Carcinoma (KIRC), using comprehensive data from The Cancer Genome Atlas (TCGA). Methods: We analyzed RNA sequencing data comprising 38,712 transcripts along with clinical variables from 534 KIRC patients. Using the preprocessing modules of the CPSM, 526 samples were retained and randomly split into training (90%) and independent test (10%) sets. CPSM's feature selection modules identified three distinct feature sets: (1) three key clinical variables (age, tumor subtype, histological grade), (2) a Prognostic Index (PI) score from nine genes selected via LASSO Cox regression, and (3) a set of univariate significant genes. Using CPSM’s model development module, we constructed four survival models: (1) Model 1 - clinical features only, (2) Model 2 - the PI score alone, (3) Model 3 - a combination of the PI score and key clinical features, and (4) Model 4 - an integration of significant univariate clinical and gene expression features. The performance of the models was evaluated using the concordance index (C-Index) and Integrated Brier Score (IBS). Results: Model 3, which integrated clinical features and the PI score, showed the best performance, achieving a C-Index of 0.79 in the training set and 0.76 in the test set, with IBS values of 0.16 and 0.23, respectively. A specific patient case (TCGA-CJ-6030-01) demonstrated that Model 3 predicted a median survival time of 73.81 months, within 5% of the actual survival time, underscoring its clinical relevance. Moreover, CPSM produced an intuitive nomogram with a C-Index of 0.76, allowing predictions of 1-year, 3-year, and 5-year survival probabilities, serving as a valuable tool for clinical decision-making. Conclusions: CPSM signifies a notable advancement in automated survival analysis by effectively integrating molecular and clinical data. The superior performance of Model 3 highlights the advantages of combining PI scores with clinical features to enhance prognostic accuracy in KIRC. This study emphasizes CPSM's potential utility in personalized oncology and establishes a foundation for future research on its application across various cancer types, including the integration of additional multi-omic data to further improve predictive performance. Citation Format: Harpreet Kaur, Pijush Das, Kevin Camphausen, Uma Shankavaram. Streamlining survival analysis with an automated machine learning pipeline: A case study in kidney renal clear cell carcinoma [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Functional and Genomic Precision Medicine in Cancer: Different Perspectives, Common Goals; 2025 Mar 11-13; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(5 Suppl):Abstract nr B040.

In the event of a large-scale radiological emergency, delivering timely medical aid to individuals receiving potentially lethal doses of radiation will result in improved survival and decreased severity of injuries. While it may be possible to reconstruct a dose estimate based on a location during the event and/or early symptoms presenting after the event, limitations with readily available information and inaccuracy of that estimate may not provide enough certainty for successful medical triage. Thus, individual biodosimetry assessments would assist medical professionals in providing prompt care to those who would benefit the most. In this study, a variety of accessible biospecimens (blood, plasma, serum, feces, saliva, and urine) from eight rhesus macaques irradiated with a single total body sublethal dose of 4 Gy of 60Co γ rays were collected before and up to 60 days after exposure for distribution to 10 different investigators' work sites for site-specific analyses. Results showing statistically significant changes in hematology parameters as well as gene, protein, and metabolite expression have since been published. Here, these results are combined and integrated with new data from microRNA (miRNA) expression in plasma samples as well as 16S rRNA sequencing and metabolomics data from fecal samples. A total of 40 unique miRNAs were significantly expressed on days 3, 6, 30, or 60. Metabolomic analysis of fecal samples found changes in multiple pathways, including steroid hormones, C18 (sex) hormones, and bile acid synthesis. Temporal changes were found in the gut microbiome for microbial abundance and richness. Finally, a retrospective view of the collective results demonstrated common overlapping pathways that were enriched from significantly altered biomarkers. This large, collaborative study from a single irradiated cohort demonstrates the utility of multiple timepoints, biospecimen types, and omics technologies that collectively identified 61 common biomarkers across 4 omics platforms that were enriched for pathways relevant to an acute radiation injury to the hematopoietic system that may aid future radiation biodosimetry efforts.

Abstract Background: Although radiation is known to modulate immune responses, it remains challenging to identify effective combinations of radiation and immune-based therapies. Because tumor, immune, and other cells release extracellular vesicles (EVs) continuously and because those vesicles are composed of lipid membrane-bound complexes of proteins and nucleic acids, we have developed a suite of tools to enable systematic subset characterization of tumor and immune EV subsets. Broadly speaking, these tools provide a new, informative approach for interrogating tumor and immune biomarkers that may be used not only for predictive and prognostic purposes, but also for adaptive treatment strategies. In this study, we hypothesized that EV subsets could be used to monitor tumor and immune responses to radiation. The objective of this study was to evaluate radiation-induced changes in tumor-cell derived EVs from cell culture and from patient serum EV subsets. Methods: For these studies, we have developed and refined a set of rigorous tools and protocols for the enumeration, labeling, and sorting of EVs individually or as subsets. We applied these methods to EVs isolated at a series of timepoints from a panel of tumor cell lines irradiated at different doses. In a parallel fashion, we also applied these methods to clinical serum samples obtained from patients before, at the completion of, and one month following radiation therapy. For all samples, EV repertoires were evaluated by multiplex EV repertoire assays, and selected EV subsets were further enriched by immunoaffinity for cargo analysis. Results: EV surface marker repertoires demonstrated both donor-specific and treatment response-specific signatures, for both tumor-associated and immune-associated EVs. For patients with bladder and colon cancer, characteristic EV tumor and immune marker signatures were noted between patients, including CD326 and CD69, and various donor-specific changes in tetraspanins (CD63, -81, and -9), CD42a, and CD62P were observed at the completion of radiation treatment. Certain EV subsets could be consistently isolated with affinity chromatography from donor serum samples but were not strongly represented in multiplex assay based on co-expression of tetraspanins. Conclusions: Our results establish feasibility for future EV subset studies, particularly in clinical trials investigating tumor and immune biomarkers for patients receiving combined radiation and immunotherapy treatments. Specifically, we demonstrate a robust method for evaluation of individual donor responses, including kinetic changes in tumor-associated markers (CD326) and immune-associated markers (CD69). Furthermore, each patient’s serum EVs demonstrated distinctive treatment-response patterns, with inter-donor differences appearing to relate to tumor-specific attributes and/or radiation parameters, such as marrow volume irradiated. Ongoing studies are further exploring EV-borne RNA signatures using a total RNAseq pipeline that is customized for EV RNA. Citation Format: Jubin George, Stephanie Chidester, Michelle L. Pleet, Kevin Camphausen, Freddy Escorcia, Jennifer C. Jones. Systematic analysis of tumor and immune EV subset changes with radiation therapy. [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Translating Targeted Therapies in Combination with Radiotherapy; 2025 Jan 26-29; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(2_Suppl):Abstract nr B022.

Introduction: Osimertinib has exhibited impressive efficacy in advanced EGFR-mutated NSCLC; however, resistance is inevitable. We hypothesized that local ablative therapy (LAT) for oligoprogressive disease (up to five sites), followed by osimertinib rechallenge, would be safe and provide additional second progression-free survival (PFS2) benefit. Methods: This prospective phase 2 trial enrolled EGFR-mutated NSCLC patients in three cohorts: tyrosine kinase inhibitor (TKI)–naive (cohort 1), previously treated with TKI and developed acquired T790M resistance mutation (cohort 2), or previously treated with osimertinib and developed resistance (cohort 3). Patients in cohorts 1 and 2 received upfront osimertinib followed by LAT on oligoprogression, followed by osimertinib rechallenge. Cohort 3 patients underwent LAT on enrollment, followed by osimertinib rechallenge. The primary end points were safety, tolerability, and PFS2 among the patients who underwent LAT across all three cohorts combined. Secondary end points were PFS1 and overall response rates. Results: A total of 37 patients with EGFR-mutated NSCLC were enrolled; 25 in cohort 1, nine in cohort 2, and three in cohort 3. A total of 21 patients received LAT across all three cohorts combined, yielding a median PFS2 of 3.7 months (95% confidence interval: 1.9–4.6 mo) for this population. A subgroup with exceptionally long PFS2 was identified that achieved lower tumor burden and circulating tumor DNA–negative minimal residual disease of the EGFR clone with osimertinib before undergoing LAT. Most adverse events related to LAT were grades 1 and 2. Conclusions: This is the first prospective trial exploring local therapy and osimertinib rechallenge on oligoprogression on osimertinib. Interim analysis revealed that PFS2 in the intention-to-treat patients did not meet its primary goal when compared with historical data on continuation of first-generation EGFR TKIs after LAT. However, definite LAT can be carefully considered in patients using circulating tumor DNA–negative minimal residual disease status as a biomarker for predicting who will benefit from continuation of osimertinib post-LAT.

Purpose Multidisciplinary tumor boards (MTBs) integrate clinical, molecular, and radiological information and facilitate coordination of neuro-oncology care. During the COVID-19 pandemic, our MTB transitioned to a virtual and multi-institutional format. We hypothesized that this expansion would allow expert review of challenging neuro-oncology cases and contribute to the care of patients with limited access to specialized centers. Methods We retrospectively reviewed records from virtual MTBs held between 04/2020–03/2021. Data collected included measures of potential clinical impact, including referrals to observational or therapeutic studies, referrals for specialized neuropathology analysis, and whether molecular findings led to a change in diagnosis and/or guided management suggestions. Results During 25 meetings, 32 presenters discussed 44 cases. Approximately half ( n = 20; 48%) involved a rare central nervous system (CNS) tumor. In 21% ( n = 9) the diagnosis was changed or refined based on molecular profiling obtained at the NIH and in 36% ( n = 15) molecular findings guided management. Clinical trial suggestions were offered to 31% ( n = 13), enrollment in the observational NCI Natural History Study to 21% ( n = 9), neuropathology review and molecular testing at the NIH to 17% ( n = 7), and all received management suggestions. Conclusion Virtual multi-institutional MTBs enable remote expert review of CNS tumors. We propose them as a strategy to facilitate expert opinions from specialized centers, especially for rare CNS tumors, helping mitigate geographic barriers to patient care and serving as a pre-screening tool for studies. Advanced molecular testing is key to obtaining a precise diagnosis, discovering potentially actionable targets, and guiding management.

Radiation-induced gastrointestinal (GI) dose constraints are still a matter of concern with the ongoing evolution of patient outcomes and treatment-related toxicity in the era of image-guided intensity-modulated radiation therapy (IMRT), stereotactic ablative radiotherapy (SABR), and novel systemic agents. Small bowel (SB) dose constraints in pelvic radiotherapy (RT) are a critical aspect of treatment planning, and prospective data to support them are scarce. Previous and current guidelines are based on retrospective data and experts’ opinions. Patient-related factors, including genetic, biological, and clinical features and systemic management, modulate toxicity. Omic and microbiome alterations between patients receiving RT to the SB may aid in the identification of patients at risk and real-time identification of acute and late toxicity. Actionable biomarkers may represent a pragmatic approach to translating findings into personalized treatment with biologically optimized dose escalation, given the mitigation of the understood risk. Biomarkers grounded in the genome, transcriptome, proteome, and microbiome should undergo analysis in trials that employ, R.T. Bioinformatic templates will be needed to help advance data collection, aggregation, and analysis, and eventually, decision making with respect to dose constraints in the modern RT era.

Abstract Gliomas exhibit nearly uniform recurrence and poor prognosis. Management of high-grade tumors is surgery followed by chemoirradiation (CRT). Radiographic progression is assessed using contrast-enhanced MRI with reporting captured in Electronic Health Records (EHR). The ability to harness large scale EHR data is limited by the Response Assessment in Neuro-Oncology (RANO) tumor progression criteria, which defines progression as an aggregate of clinical and/or radiographic parameters requiring clinician judgement. As a result, glioma progression is not captured systematically in large data sets, limiting Progression Free Survival (PFS) as an outcome endpoint in data analysis. We developed an AI-based method using natural language processing to capture PFS parameters for analysis of MRI radiology reports. 1088 available brain MRI radiology reports for 81 patients with a pathologically confirmed diagnosis of glioblastoma (GBM) were aggregated in the NIH Integrated Data Analysis Platform. MRI reports were systematically analyzed and PFS manually captured using RANO criteria as ground truth. Common report terms indicating progression were compiled and included in task prompts applied to Large Language Model (LLM) Extraction Tools. The words progression (n=1047) and stable (n=1133) did not necessarily indicate either overall progression or stability in a report. Yet, in the 24 (30%) patients who progressed within 3 months of CRT, recurrence and mass effect occurred in 3.7% and 53% of reports, respectively, which was comparable to the term frequencies of 5.8% and 54% in patients with stable disease. Thus, mass effect and recurrence could not necessarily be used alone as terms to signal overall progression. The analysis identified other commonly employed radiology report terms relevant to determining tumor progression including enhancement (n=2445), perfusion (n=2085), enhancing (n=1785), resection (n=1576), and increased (n=1331), which allowed for experimentation with more detailed LLM extraction task prompts. Our method flagged radiographic progression prior to clinician-coded progression in 43 (53%) patients. Future directions include handling further clinical context such as surgical features, radiation therapy dates, progress notes, and the coadministration of agents such as bevacizumab and steroids. These results indicate the feasibility of LLMs to identify tumor progression dates using EHR text with potential transferability to large glioma data sets pending further optimization and validation. Citation Format: Shreya Chappidi, Hawon Lee, Sarisha Jagasia, Casey Syal, George Zaki, Dylan Junkin, Nathan Golightly, Patrick Chitwood, Kevin Camphausen, Andra Krauze. Defining and capturing progression in glioma by harnessing NLP in unstructured electronic health records [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6199.

Glioblastoma (GBM) is a primary central nervous system malignancy with a median survival of 15–20 months. The presence of both intra- and intertumoral heterogeneity limits understanding of biological mechanisms leading to tumor resistance, including immune escape. An attractive field of research to examine treatment resistance are immune signatures composed of cluster of differentiation (CD) markers and cytokines. CD markers are surface markers expressed on various cells throughout the body, often associated with immune cells. Cytokines are the effector molecules of the immune system. Together, CD markers and cytokines can serve as useful biomarkers to reflect immune status in patients with GBM. However, there are gaps in the understanding of the intricate interactions between GBM and the peripheral immune system and how these interactions change with standard and immune-modulating treatments. The key to understanding the true nature of these interactions is through multi-omic analysis of tumor progression and treatment response. This review aims to identify potential non-invasive blood-based biomarkers that can contribute to an immune signature through multi-omic approaches, leading to a better understanding of immune involvement in GBM.

In the original publication [...].

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

Purpose/Objective(s) Glioblastoma (GBM) is characterized by poor survival outcomes and high rates of recurrence due to rapid proliferation and treatment resistance. The CXCL family of chemokines has been implicated in multiple aspects of GBM biology, including angiogenesis and tumor progression. There is a pressing need to identify clinically relevant, easily measurable biomarkers to further the understanding of treatment response. CXCL levels in GBM patients before and after chemoirradiation therapy (CRT) were analyzed for alteration in serum and relationship to clinical and radiation therapy (RT) data. Materials/Methods Serum samples from 109 patients with pathologically proven GBM (diagnosed 2005-2023) were collected before and after the completion of CRT and analyzed using the Somalogic 7k proteomic panel. 21 members of the CXCL family were identified, 14 of which were unique. CXCL levels were linked to clinical and RT data. Statistical analyses (Wilcoxon test, Spearman r correlation, Kaplan Meier) were performed to explore alterations following treatment and associations with clinical features, overall survival (OS), and progression free survival (PFS). Results Eight unique CXCLs were found to be significantly altered with CRT: CXCL2 (P = 0.0002), CXCL3 (P<0.0001), CXCL5 (P<0.0001), CXCL6 (P = 0.023), CXCL10 (P<0.0001), CXCL11 (P<0.0001), CXCL13 (P = 0.014), and CXCL16 (P<0.0001). CXCL2, 3, 5, 6, and 11 decreased while CXCL10, 13, and 16 increased in response to CRT. Statistically significant interactions were noted between chemokines and between clinical variables. CXCL1, 2, 3, 5, and 6 were strongly directly correlated with each other (P<0.05) and were weakly inversely correlated with age (CXCL1, 2, 3). CXCL16 alteration following CRT was inversely correlated with the alteration of CXCL1, 2, 3, 6, and 13 while being directly correlated with the alteration of CXCL8 and 10 (P<0.05). OS and PFS were strongly associated with age, MGMT status, and GTVT1 (P<0.05). CXCL16 was associated with MGMT status with MGMT methylated patients exhibiting more significant increases in serum CXCL16 (P = 0.026) pre vs. post CRT. Elevated CXCL16 was associated with improved OS (P = 0.031) (median OS 27 months increased levels vs. 18 months for lower or decreased levels) but was not associated with PFS. Pre CRT CXCL5 had the strongest direct correlation with GTVT1 (P = 0.005) and GTVT2 (P = 0.008) while pre–CRT CXCL13 was directly correlated with GTVT2 only (P = 0.006); however, neither was associated with OS or PFS. Conclusion Several members of the CXCL family are measurable in serum and significantly altered in response to CRT in GBM patients. The directionality of these changes and further analysis with clinical and RT data could enhance the understanding of molecular signaling pathways relevant to the GBM response to CRT. Association of CXCL16 with MGMT status may indicate potential for these molecules to be employed as biomarkers in GBM.

Introduction Patient selection remains challenging as the clinical use of re-irradiation (re-RT) increases. Re-RT data are limited to retrospective studies and small prospective single-institution reports, resulting in small, heterogenous data sets. Validated prognostic and predictive biomarkers are derived from large-volume studies with long-term follow-up. This review aims to examine existing re-RT publications and available data sets and discuss strategies using artificial intelligence (AI) to approach small data sets to optimize the use of re-RT data.

IntroductionPatient selection remains challenging as the clinical use of re-irradiation (re-RT) increases. Re-RT data are limited to retrospective studies and small prospective single-institution reports, resulting in small, heterogenous data sets. Validated prognostic and predictive biomarkers are derived from large-volume studies with long-term follow-up. This review aims to examine existing re-RT publications and available data sets and discuss strategies using artificial intelligence (AI) to approach small data sets to optimize the use of re-RT data.MethodsRe-RT publications were identified where associated public data were present. The existing literature on small data sets to identify biomarkers was also explored.ResultsPublications with associated public data were identified, with glioma and nasopharyngeal cancers emerging as the most common tumor sites where the use of re-RT was the primary management approach. Existing and emerging AI strategies have been used to approach small data sets including data generation, augmentation, discovery, and transfer learning.ConclusionsFurther data is needed to generate adaptive frameworks, improve the collection of specimens for molecular analysis, and improve the interpretability of results in re-RT data.

Background Glioblastomas (GBM) are rapidly progressive, nearly uniformly fatal brain tumors. Proteomic analysis represents an opportunity for noninvasive GBM classification and biological understanding of treatment response. Purpose We analyzed differential proteomic expression pre vs. post completion of concurrent chemoirradiation (CRT) in patient serum samples to explore proteomic alterations and classify GBM by integrating clinical and proteomic parameters. Materials and methods 82 patients with GBM were clinically annotated and serum samples obtained pre- and post-CRT. Serum samples were then screened using the aptamer-based SOMAScan® proteomic assay. Significant traits from uni- and multivariate Cox models for overall survival (OS) were designated independent prognostic factors and principal component analysis (PCA) was carried out. Differential expression of protein signals was calculated using paired t-tests, with KOBAS used to identify associated KEGG pathways. GSEA pre-ranked analysis was employed on the overall list of differentially expressed proteins (DEPs) against the MSigDB Hallmark, GO Biological Process, and Reactome databases with weighted gene correlation network analysis (WGCNA) and Enrichr used to validate pathway hits internally. Results 3 clinical clusters of patients with differential survival were identified. 458 significantly DEPs pre- vs. post-treatment, 316 upregulated, 142 downregulated emerged including several pathways relevant to cancer metabolism and progression. The worst survival group (median OS 13.2 months) was associated with DEPs affiliated with proliferative pathways and distinct oppositional response (including RT) as compared to better-performing groups (intermediate, median OS 22.4 months; highest, median OS 28.7 months). Opposite signaling patterns across multiple analyses in several pathways (notably fatty acid metabolism, TNFα via NF-κB, Myc target V1 signaling, UV response, unfolded protein response, peroxisome, and interferon response) were distinct between clinical survival groups and supported by WGCNA. 9 proteins were statistically signficant for OS with 1 (CEACAM16) supported by KM. Conclusion Distinct proteomic alterations with hallmarks of cancer, including progression, resistance, stemness, and invasion, were identified in serum samples obtained from GBM patients pre vs. post CRT and corresponded with clinical survival. The proteome can potentially be employed for glioma classification and biological interrogation of cancer pathways.

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