Citations

Abstract Objective Interictal epileptiform spikes (IESs) are widely used biomarkers of epileptogenic tissue in presurgical epilepsy evaluation. However, their reliability for seizure onset zone (SOZ) localization in high‐density intracranial electroencephalography (iEEG) remains inconsistent. It is underexplored whether this variability reflects methodological differences across detection algorithms or a limitation of IESs as a biomarker. This study evaluates whether IES detection alone provides sufficient spatial specificity for accurate SOZ localization in iEEG. Methods We systematically benchmarked eight established IES detection algorithms using iEEG recordings from 35 patients with drug‐resistant focal epilepsy. To provide a clinically meaningful reference, algorithm performance was evaluated against the clinically defined SOZ. To refine detection outputs and move beyond only IES detection, we applied masking strategies to isolate informative IES subsets based on rate, amplitude, frequency, and co‐occurrence with high‐frequency oscillations (HFOs). Conceptually, masking acts as a spatial refinement strategy that progressively restricts diffuse IES activity to more focal, pathologically relevant event populations. Results Despite substantial variability in detection characteristics, SOZ localization performance (ratio of events originating from SOZ divided by total number of events) converged within a narrow range (.23–.51), suggesting that IESs exhibit an intrinsic ceiling in SOZ localization performance when used alone. Amplitude‐based masking yielded modest improvements, whereas frequency‐based masking reduced performance. In contrast, IES‐HFO co‐occurrence markedly improved localization, particularly when combined with rate‐based masking, achieving a median SOZ localization ratio of .89, exceeding IESs (.37) and HFOs (.63) evaluated independently. Significance These findings demonstrate that IES‐based SOZ localization may require moving beyond detection‐only approaches. The integration of informed, biomarker‐driven masking strategies provides a more robust framework for identifying the activity most relevant to the epileptogenic network and paves the way for future incorporation into the clinical decision‐making pipeline.

Abstract Background Delirium is often underdiagnosed in clinical practice and is not routinely coded for billing. While manual chart review can identify delirium, it is labor-intensive and impractical for large-scale studies. Natural language processing (NLP) can analyze unstructured text in electronic health records (EHRs) to extract meaningful clinical information. Methods To support national integration of NLP for EHR-based delirium identification across different institutions, we launched the Delirium Interest Group within the national Evolve to Next-Gen Accrual to Clinical Trials (ENACT) NLP Working Group. This paper outlines our initial efforts to standardize, evaluate, and translate an NLP-based delirium detection model into the i2b2/ENACT platform. Results Multisite contextual inquiry identified several key challenges, including variations in local screening practices (eg, tools used, documentation frequency, and quality control), the need for harmonized definitions in the context of EHRs, and the complexity of modeling temporal logic. Multisite NLP evaluation revealed variable performance degradation driven by differences in delirium screening practices, clinical documentation patterns and semantics, and note syntactic structures. Conclusion Our work represents an important first step toward enabling scalable and standardized NLP-based delirium detection across institutions. By engaging diverse institutions through the ENACT NLP Working Group, we identified shared challenges and site-specific variations that impact model implementation and performance. Our collaborative approach enabled the development of a more robust framework for delirium identification across heterogeneous EHR systems. Future efforts will build on this foundation to enhance the validity, usability, and translational impact of delirium detection.

Abstract Error analysis is a critical step in evaluating and improving clinical concept extraction models, the most common clinical natural language processing (NLP) task. Unlike corpus annotation, which follows standardized protocols for creating gold-standard datasets, error analysis requires nuanced judgment grounded in both clinical expertise and NLP knowledge. This is especially important given the heterogeneity of clinical text, where variations in documentation style, note structure, and terminology can substantially influence model behavior. Despite its importance, there is currently no standardized, user-level framework to support systematic error analysis in clinical concept extraction task. In this study, we developed and validated MedError, a machine-assisted, human-in-the-loop framework designed to standardize and enhance error analysis for clinical concept extraction tasks. We collected and manually curated a corpus of 1,187 unique errors from a total of 4,237 notes across three different distinct hospitals. The error categories were defined using our previously validated error taxonomy and included 480 false negatives and 707 false positives across 25 error types and 48 clinical concept categories. We evaluated the performance of three proprietary and three open-source large language models (LLMs) in automatically classifying these errors into 26 and 15 predefined categories. We further developed a machine-assisted framework, MedError, which integrates best practices in error analysis, LLM-assisted classification and reasoning, and a user-friendly interface to enable more efficient, reproducible, and context-aware error analysis. The framework supports both single-site and federated multisite error analysis, facilitating the effective translation of clinical NLP systems into real-world settings.

Abstract Background The ataxia-telangiectasia mutated (ATM) protein coordinates the cellular response to therapeutic radiation, and ATM inhibitors can potentially enhance the efficacy of radiation in otherwise radiation-resistant tumors. Methods The small molecule ATM kinase inhibitor WSD0628 was specifically designed for enhanced distribution across the blood-brain barrier to more effectively treat glioblastoma (GBM) and brain metastases in combination with radiation therapy (RT). GBM and brain metastasis patient-derived xenograft models were used to understand target inhibition, radiosensitization, inhibition of the DNA damage response, and in vivo efficacy. Results Initial in vitro characterization of WSD0628 demonstrates a high level of selectivity across kinase families, limited aldehyde oxidase liability, and low risk of hERG interactions. Consistent with a central role for ATM in radiation response, WSD0628 blocked radiation-induced signaling and enhanced radiosensitivity in U251 glioma cells and brain tumor PDXs GBM120 and M12. In comparison to control or RT alone in orthotopic PDXs, the combination of WSD0628 with RT markedly prolonged median survival—GBM12 (19, 55, and 408 days, respectively); GBM43 (26, 44, and 143 days, respectively); GBM120 (51, 89, and 231 days, respectively); M12 (17, 39, and 190 days, respectively). Pharmacokinetic and pharmacodynamic testing after treatment in orthotopic GBM43 tumors showed inhibitory levels of WSD0628 and a reduction of γH2AX foci in the combination-treated tumors. Conclusions Collectively, these results suggest a promising role for WSD0628 in combination with RT in brain tumors and provide the rationale for an ongoing Phase 0/1A clinical trial testing this combination in recurrent GBM.

Abstract Glioblastoma (GBM) remains one of the most therapy-resistant malignancies with frequent local failures despite aggressive surgery, chemotherapy, and ionizing radiation (IR). Small molecule inhibitors of DNA-dependent protein kinase (DNA-PKi) are potent radiosensitizers currently in clinical trials. Determining which patients may benefit from radiosensitization with DNA-PKi is critical to avoid unnecessary increased risk of normal tissue toxicity. In this study, we used GBM patient-derived xenografts (PDX) in orthotopic murine models to study the relationship between molecular features, pharmacokinetics, and the radiosensitizing potential of the DNA-PKi peposertib. We show that peposertib radiosensitizes established and PDX GBM lines in vitro at 300 nmol/L and above, with a significant increase in radiosensitization by maintaining post-IR exposure for >12 hours. Radiosensitization by peposertib is mediated by catalytic inhibition of DNA-PK, and knockdown of DNA-PK by short hairpin RNA (shRNA) largely abolished the radiosensitizing effect. Peposertib decreased auto-phosphorylation of DNA-PKcs after IR in a dose-dependent manner with a delay in resolution of γH2AX foci at 24 hours. The addition of peposertib to IR significantly increased survival in GBM120 orthotopic xenografts, but not in GBM10. There was no difference in plasma or average tumor concentrations of peposertib in the two cohorts. Although the mechanism underpinning this discordant effect in vitro versus in vivo is not clear, there was an association for greater sensitization in TP53 mutant lines. Transfection of a dominant-negative TP53 mutant in baseline TP53 wild-type GBM lines significantly delayed growth and decreased nonhomologous end joining efficiency (but not homologous recombination), after peposertib exposure. See related commentary by Buchsbaum, p. 840

Abstract Background: In breast cancer, the spatial relationship between immune and tumor cells is increasingly recognized as critical in determining the efficacy of neoadjuvant therapy (NAT). This study investigated how the distribution of immune cells, particularly CD8+ T cells, relative to tumor cells affects treatment outcomes. Methods: Biopsies from 104 breast cancer patients were examined, both pre- and post-NAT, using multiplex immunohistochemistry (mIHC). This study focused on the density and distribution of CD8+ T cells, CD68+ macrophages, FoxP3+ regulatory T cells, and other immune markers. Results: The spatial proximity of CD8+ T cells within 20 µm of cancer cells (N20-CD8+T) was strongly correlated with improved clinical outcomes across diverse tumor subtypes and NAT regimens. A significant post-NAT decrease in CD4+, CD68+, and FoxP3+ cells highlights notable shifts in the tumor immune landscape. A noteworthy finding was the positive correlation between CXCL9 expression and N20-CD8+T cells, suggesting a mechanism for immune cell recruitment and interaction within the tumor microenvironment. These results highlight the complex interplay among NAT, immune cell modifications, and patient outcomes. Conclusions: This study revealed that the proximity of N20-CD8+ T cells to tumor cells is consistently associated with favorable NAT outcomes, transcending tumor subtypes, and treatment differences. This highlights the potential of immune cell dynamics, especially CD8+ T cells, as predictive biomarkers for therapeutic responses in breast cancer, underscoring their vital role in tumor immunity.

Abstract Glioblastoma (GBM) is an aggressive and highly infiltrative brain tumor with poor prognosis. Radiation therapy induces DNA damage, and DNA damage response (DDR) pathways modulate resistance to radiation therapy. Inhibiting DDR pathways by targeting ATM, a key enzyme involved in the repair of DNA double strand breaks, leads to radiosensitization and enhances cytotoxicity. Clinical translation of radiosensitizers for GBM may be influenced by a lack of effective drug delivery across the blood brain barrier (BBB). The purpose of this study is to determine the key preclinical pharmacokinetic parameters and mechanisms that influence the CNS delivery of WSD0628, a novel ATM inhibitor. In vitro clonogenic survival assays using U251 GBM cells indicate that WSD0628 shows significant radiosensitization, with a >70% decrease in surviving colonies in the presence of 10nM or higher drug concentrations in combination with 2.5 to 10 Gy radiation as compared to radiation alone. We evaluated the systemic pharmacokinetics and CNS distribution of WSD0628 in FVB wild-type mice and FVB mice lacking Pgp and Bcrp (Mdr1a/b−/−Bcrp1−/−, triple knockout) after a 1, 5, 10 mg/kg intravenous (IV) or oral (PO) dose. Clearance decreased with increase in dose and a greater than dose proportional increase in exposure was observed. WSD0628 is rapidly absorbed and is bioavailable after oral administration. Tissue distribution coefficients were determined as the ratio of exposure in tissue to plasma (Kp = AUCtissue/AUCplasma). Kp,brain was 0.22 ± 0.05, and Kp, spinal cord was 0.30 ± 0.09 in the wild-type mice. For wild-type mice, Kp,uu,brain was 0.20 ± 0.05 and Kp,uu,spinal cord was 0.26 ± 0.08, determined by correcting for binding to CNS tissues and plasma. The brain distribution of WSD0628 was only modestly increased (1.4 times) in the triple knockout mice with Kp,brain of 0.32 ± 0.06 and Kp,spinal cord of 0.41 ± 0.08. Organ distribution of WSD0628 at steady-state was studied after intraperitoneal implantation of Alzet osmotic pumps primed to release WSD0628 at 7.5 ug/hr for 25 hours. Tissue distribution coefficients, determined by organ-to-plasma steady-state concentration ratios, were highest for liver and kidneys, followed by heart, lungs and CNS. Survival studies conducted in mice bearing orthotopic GBM patient derived xenografts show a significant improvement in survival in mice receiving a combination of WSD0628 and radiation as compared to radiation alone. In conclusion, WSD0628 is a robust radiosensitizer and has moderate brain penetration that is not limited by the efflux transporters Pgp and Bcrp. Based on preliminary efficacy studies in mice, WSD0628 shows potential for treatment of GBM tumors. Evaluations are underway to determine tumor free concentrations and quantification of corresponding biomarkers. Combining this knowledge of PK of WSD0628 with the dynamics of the targeted biological pathway will inform a PK-PD-efficacy model and predict effective dosage regimens for future preclinical and clinical trials. Citation Format: Sneha Rathi, Ann Mladek, Zhiyi Xue, Ju-Hee Oh, Wenjuan Zhang, Wenqiu Zhang, Jiayan Le, Jann S. Sarkaria, William F. Elmquist. Systemic and CNS distributional pharmacokinetics of WSD0628, a novel radiosensitizer: Implications for the treatment of brain tumors [abstract]. In: Proceedings of the AACR Special Conference on Brain Cancer; 2023 Oct 19-22; Minneapolis, Minnesota. Philadelphia (PA): AACR; Cancer Res 2024;84(5 Suppl_1):Abstract nr B028.

Abstract Radiation therapy (RT) is an important non-surgical clinical treatment for glioblastoma (GBM) and brain metastases, but inherent radio-resistance can limit the efficacy in many patients. The Ataxia Telangiectasia Mutated (ATM) protein critically regulates radiotherapy-mediated DNA damage repair pathways, and defects in this kinase can lead to profound radiosensitivity. In this study, the efficacy of the highly potent ATM inhibitor WSD0628 was evaluated in GBM and melanoma brain metastasis patient derived xenografts (PDXs). In short term explant cultures of three PDXs, WSD0628 robustly suppressed RT-induced autophosphorylation of ATM at serine 1981 and ATM-mediated phosphorylation of Chk2 at threonine-68 and KAP1 at serine-824, with maximal inhibition at 100 nM. Similarly, RT-induced gH2AX foci were significantly reduced when combined with WSD0628 in the GBM43 cells. Consistent with the importance of ATM in the DNA damage response after radiation, WSD0628 significantly increased the radiosensitivity of U251 cells in clonogenic survival assays (0nM vs. 30nM at 5Gy, p<0.01). To evaluate the optimal duration of drug exposure, WSD0628 was removed from U251 cells at various intervals following irradiation with 5 Gy. Using clonogenic survival as a readout, a 10-fold increase in cytotoxicity was observed with an 8 hour drug exposure, with progressively increasing cytotoxicity gains with exposures up to 24 hours. In vivo efficacy of WSD0628 was evaluated in multiple intracranial PDX models. In an initial dose-ranging study, robust radiosensitization was observed in GBM43 treated with 2 Gy x 5 fractions combined with 5 mg/kg WSD0628 (20 day survival extension) or 10mg/kg WSD0628 (39 day with 10 mg/kg; p<0.01 for both dose levels). Moreover, a single dose of 12.5 Gy combined with 10 mg/kg WSD0628 had profound impact on treatment efficacy in the melanoma brain metastasis M12 PDX: at 180 days post-treatment, all mice in the combination group were electively euthanized, while median survival for sham or RT alone was 17d and 49d, respectively. Histologic analysis identified large intracranial tumors in all sham and RT-only treated mice, but only a small accumulation of melanotic cells without obvious tumor in the combination-treated mice. Immunohistochemical staining of NeuN and GFAP in the combination-treated mice showed preserved neuronal density at the 180-post treatment timepoint and minimal reactive gliosis within the ‘tumor scar’. An ongoing study is comparing this single-fraction 12.5 Gy regimen to a 2.4 Gy x 10 fraction regimen alone or in combination with WSD0628. Collectively, this study demonstrates the potential for profound radiosensitizing effects of WSD0628 in combination without obvious neuronal toxicity and has provided the scientific rational for the first-in-man study of this combination in recurrent GBM (NCT05917145) at Mayo Clinic. Citation Format: Zhiyi Xue, Ann C. Mladek, Sneha Rathi, Danielle M. Burgenske, Shiv K. Gupta, Brett L. Carlson, Zeng Hu, Lauren L Ott, Katrina K. Bakken, Rachael A. Vaubel, William F Elmquist, Wei Zhong, Jann N Sarkaria. The radiosensitizing effects of the novel brain penetrant and potent ATM inhibitor WSD0628 in glioblastoma and melanoma patient derived xenografts [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr A148.