Citations

e15509 Background: Large language models (LLMs) are increasingly explored for oncology decision support, yet comparative evidence evaluating medical-specific versus general-purpose LLMs using expert-reviewed clinical cases is limited. We compared a medical-specific LLM (Google MedGemma 27B) with a general-purpose LLM (ChatGPT-5) using gastrointestinal (GI) and genitourinary (GU) oncology vignettes. Methods: This cross-sectional study evaluated 32 de-identified oncology vignettes (26 GI, 6 GU) created by a U.S. board-certified medical oncologist and informed by real-world cases. Vignettes included text-only information on the clinical context, description of pathology, tumor staging, biomarkers, and relevant comorbidities. Both LLMs were prompted to provide National Comprehensive Cancer Network(NCCN) guideline-based treatment recommendations, monitoring strategies, maintenance therapy considerations, and references, along with a self-reported confidence score (0–100). Board-certified U.S. medical oncologists independently reviewed outputs using a structured instrument; two disease specific expert oncologists evaluated each case. Outcomes included correctness and comprehensiveness (1–10 scales), hallucinations (factually fabricated or unsupported statements), NCCN traceability, potential patient harm, and clinician trust. Associations between accuracy and comprehensiveness were assessed using correlation coefficients. Inter-rater reliability was assessed using a two-way random-effects intraclass correlation coefficient (ICC). Results: ChatGPT-5 demonstrated higher mean correctness (8.65 vs 4.68) and comprehensiveness (8.68 vs 5.22) scores compared with MedGemma 27B. Accuracy and comprehensiveness were strongly correlated overall (Pearson r = 0.93, p < 0.001), with significant model-specific correlations (ChatGPT-5 r = 0.86, p < 0.001; MedGemma r = 0.77, p < 0.001). Hallucinations occurred less frequently with ChatGPT-5 (12.3% vs 65%), and NCCN-traceable recommendations were more common (75.4% vs 36.7%). Responses judged as potentially harmful occurred in 18.5% of ChatGPT-5 outputs versus 70% of MedGemma outputs. Clinician trust favored ChatGPT-5 (84.6% vs 15%). Inter-rater reliability improved with rater averaging (ICC(A,k): correctness 0.72; comprehensiveness 0.64). Conclusions: In expert-reviewed GI and GU oncology cases, ChatGPT-5 demonstrated higher clinical accuracy, comprehensiveness, and guideline alignment than a medical-specific LLM; however, a substantial rate of hallucinations and potentially harmful recommendations persisted across models. These findings underscore that, despite performance differences, current LLMs require rigorous clinician oversight and should not be used as standalone decision-support tools in oncology.

Abstract Aims To characterize global variation in cardiac rehabilitation (CR) staffing, indications, and prescribed dose by country income classification and compare findings with the International Council of Cardiovascular Prevention and Rehabilitation’s 2016 Global Audit. Methods and results Cross-sectional REDCap survey of Phase II/post-discharge CR programmes (May–September 2025). Outcomes were multidisciplinary team size/composition, accepted cardiac and non-cardiac indications, and centre-based CR dose. Comparisons were made across World Bank country income classifications and Audit decade (2016 vs. 2025). Of 7025 programmes identified globally, 1505 (median programme response rate = 62%/country) initiated a survey, from 90/113 (80%) countries with CR. CR was delivered by a multidisciplinary team of 7 (median; P25–P75 = 5–8; nurses, physiotherapists, cardiologists, and dietitians most common); teams were smaller in low-income countries (median = 4; P &lt; 0.001), and physician specialty varied. Non-low-income programmes consistently accepted (≥90%) patients with acute coronary syndrome, with programmes in low-income countries less often accepting related interventions (e.g. surgery: low-income 0% vs. high-income 97%; P &lt; 0.01), but non-cardiac indications were more frequently accepted (e.g. primary prevention, diabetes, and lung disease). Acceptance of heart failure increased compared with the previous Audit (70% vs. 87%; P &lt; 0.001). Centre-based programmes prescribed a median of 20 h (P25–P75 = 12–30), typically two sessions/week over 8 weeks. Globally, 65 (72%) countries prescribed a median ≥12 sessions/programme. Programmes in upper-middle-income countries prescribed more sessions (median = 24, P25–P75 = 12–36; P &lt; 0.001). Dose did not differ meaningfully between Audits (P = 0.07). Conclusion CR programmes globally are offered by small, but multidisciplinary teams. They accept guideline-indicated patients, with applicable variation by context. Insufficient dose in one-quarter of countries underscores investment need.

8049 Background: Multiple pulmonary nodules (MPN) pose significant challenges for accurate staging of non-small cell lung cancer (NSCLC). As next generation sequencing (NGS) technologies have become more widely available on smaller amounts of tissue, the prevalence and clinical implications of heterogeneity between MPNs are poorly understood. This study retrospectively evaluates the histopathological and NGS findings after MPN biopsy in a single robotic-assisted bronchoscopy (RaB) procedure. Methods: 191 patients were retrospectively enrolled at 6 geographically diverse centers in the US. All patients had 2+ pulmonary nodules biopsied and staging EBUS in the same procedure. An expert molecular pathologist reviewed NGS data between nodule pairs to determine if they were discordant, concordant, or indeterminate. Results: Of 191 patients, 85 patients had 2 malignant nodules with 15 of those demonstrating histopathologic heterogeneity (e.g., one adenocarcinoma, one squamous cell carcinoma). 26 patients with 2 malignant nodules of similar histopathology demonstrated molecular heterogeneity between nodules. These cases were reviewed by an expert molecular pathologist who determined the following breakdown: -15 discordant pairs (57.7%) -3 concordant pairs (11.5) -8 indeterminate pairs (30.8%) Breakdown by nodule density: -Discordant: 7 same density, 8 different density -Concordant: 3 same density -Indeterminate: 5 same density, 3 different density Breakdown by nodule laterality: -Discordant: 7 ipsilateral, 8 contralateral -Concordant: 2 ipsilateral, 1 contralateral -Indeterminate: 4 ipsilateral, 4 contralateral Staging EBUS was performed in all 26 cases with malignancy identified in 2 cases. NGS sufficiency for NSCLC from three sites with standardized NGS reporting was as follows: 1 st nodule biopsied: 42/54 (77.8%), 2 nd nodule biopsied: 37/51 (72.5%), 3 rd nodule biopsied: 2/4 (50%). In total, 90/109 (82.6%) NSCLC samples had adequate tissue for NGS. In 85 cases with 2 malignant nodules, 35 (41.2%) had all nodules Stage IIB or lower. Notably, 25 (29.4%) of the cases did not have NGS available on one or more of the nodules, thus precluding staging. Conclusions: This study highlights the frequency and clinical relevance of histopathologic and molecular heterogeneity among synchronous MPNs. Importantly, heterogeneity was not limited to contralateral or anatomically distant lesions, challenging the assumption that spatial proximity implies clonal relatedness. Further, this study shows that relying solely on density or radiographic features of a nodule is an incomplete discriminator between separate primary lung cancers and intrapulmonary metastasis. Our findings support the need for an integrated diagnostic paradigm in which sampling of multiple lesions is used to inform clonality, refine staging, and guide precision therapy.

5069 Background: Inhibition of androgen receptor (AR)-signaling results in increases in double-strand DNA breaks and genomic instability. PARP inhibitors (PARPi) are a standard treatment option for patients (pts) with homologous recombination repair mutations (HRRm). Combinations of PARPi and AR pathway inhibitors (ARPI) are associated with synthetic lethality in multiple pre-clinical studies. Clinical trials have suggested potential benefits with co-targeting PARPi and ARPI in pts with and without HRRm. Methods: Alliance A031902 compared radiographic progression-free survival (rPFS) and overall survival (OS) as co-primary endpoints with enzalutamide and rucaparib (Enz/Ruca) versus enzalutamide plus placebo (Enz/Plac) for pts with metastatic castration resistant prostate cancer (mCRPC) regardless of BRCA/HRRm status. This trial was conducted by the Alliance for Clinical Trials in Oncology in 36 centers in the US. Target sample size was 1,002 patients. Eligibility required mCRPC with or without measurable metastatic disease and was agnostic regarding HRRm status. The trial was discontinued in 10/2023 due to changes in the treatment landscape for this pt population. Results: Between 10/21 and 4/23, 61 pts were randomized to Enz/Ruca (n=30) or Enz/Plac (n=31). Known BRCA1/2 or PALB2 mutations were present in 1 pt (3.3%) in the Enz/Ruca arm and 1 pt (3.2%) in the Enz/Plac arm; wildtype in 14 (46.7%) vs 15 (48.4%), and unknown in 15 (50%) vs 15 (48.4%), respectively. Three patients (11.5%) in the Enz/Ruca arm discontinued due to an adverse event (AE) vs 0 in the Enz/Plac arm. Among response-evaluable pts, 7/26 (27%) in the Enz/Ruca arm discontinued due to disease progression vs 10/31 (33%) in the Enz/Plac arm. Most common grade 3+ AEs were anemia (33% vs 3%) and hypertension (17% vs 10%) in the Enz/Ruca arm and Enz/Plac arm, respectively. All 61 pts were evaluable for response assessment. PSA declines ≥50% and ≥90% were observed in 79% vs 45% and 60% vs 43% of pts receiving Enz/Ruca versus Enz/Plac, respectively. Median PFS in the Enz/Ruca vs Enz/Plac arms was 17.1 vs 14.2 mos, respectively (HR 1.13 [95% CI: 0.42-3.01], p=0.81), median rPFS (inclusive of unequivocal clinical progression) was 17.1 vs 11.7 mos ( HR 0.72; 95% CI: 0.31-1.67, p = 0.44), and median OS was not reached vs 27.0 mos (HR 0.65; 95% CI: 0.29-1.45, p=0.29). Conclusions: In this prematurely terminated phase III study, Enz/Ruca was generally well tolerated, with the safety profile consistent with prior PARPi/ARPI combinations. Efficacy outcomes for rPFS and OS are inconclusive due to limited sample size. Definitive conclusions regarding the clinical benefit of this combination cannot be determined from this study. Clinical trial information: NCT04455750 .

5004 Background: Patients (pts) with metastatic hormone-sensitive prostate cancer (mHSPC) treated continuously with testosterone suppression (TS) plus an androgen receptor pathway inhibitor (ARPI) can experience cumulative toxicity. We conducted this single-arm phase 2 trial (NCT05241860) to test the hypothesis that pts who achieve favorable response to TS + ARPI can have prolonged treatment-free interval with testosterone recovery after treatment interruption (TI). Here we report on the primary endpoint. Methods: Eligible pts had mHSPC by conventional imaging with prostate-specific antigen (PSA) ≥ 5 ng/ml and testosterone ≥ 150 ng/dl prior to starting TS+ARPI, with PSA < 0.2 ng/ml and testosterone < 50 ng/dl at the time of enrollment after having received TS for 540-750 days and ARPI for ≥ 360 days. Prior local therapy, radiation to metastases and docetaxel were permitted. After enrollment, pts discontinued TS and ARPI and were followed with PSA and testosterone levels every 3 months (mo), CT/MRI and bone scan at least every 6 mo, and FACT-P questionnaire for patient-reported outcomes (PROs) every 6 mo. Treatment was resumed for PSA ≥ 5 ng/ml, radiographic change (progressive disease [PD] per RECIST 1.1 on CT/MRI or unconfirmed PD per PCWG3 on bone scan), or prostate cancer-related symptoms. The primary endpoint was proportion of men remaining treatment-free 18 mo after TI with eugonadal testosterone ( > 150 ng/dl). Target enrollment was 75 pts to differentiate 18-mo treatment-free rates of 0.30 (H 0 ) and 0.45 (H a ). Results: Of 79 pts enrolled between 07/2022 and 03/2024, 78 were eligible and underwent TI. Among these 78, median PSA prior to starting TS+ARPI was 19 (range 5-6759), 27 (35%) had high volume disease, 31 (40%) never received local therapy, and 55 (71%) did not receive radiation to metastases. By 18 mo after TI, 67% (52/78) recovered testosterone and 58% (45/78) remained treatment-free; 41% (32/78) remained treatment-free with testosterone recovery (80% CI 33.5-48.9%, one-sided p 0.0249). At median follow-up of 21.2 mo, 35% (27/78) resumed initial TS+ARPI after meeting re-initiation criteria (of whom 1 required treatment switch for PD 9 mo later); 5% (4/78) resumed prior to meeting criteria; 9% (7/78) pursued alternative therapy instead of resuming TS+ARPI per protocol; 5% (4/78) withdrew; 1 died of myocardial infarction prior to resuming treatment. 4 pts died, 1 of prostate cancer. Conclusions: The primary objective was achieved with 41% of favorable responders to TS+ARPI remaining treatment-free with testosterone recovery at 18 mo after TI. Analyses of biomarkers and PROs are in progress, and pts will be followed for long-term outcomes. Support: U10CA180821, U10CA180882, UG1CA189823, https://acknowledgments.alliancefound.org, Veracyte Inc. Clinical trial information: NCT05241860 .

4559 Background: Baseline QOL has been linked to survival in mRCC therapy clinical trials (Cella, 2024); however, there is a paucity of prospectively collected real-world QOL data across front-line management strategies. We used data from ODYSSEY to examine the relationship between baseline QOL and early death. Methods: ODYSSEY is a multi-site, prospective observational study of 468 US pts with mRCC. Pts must have mRCC (any histology); have no prior systemic therapy (ST) or be within 3 months of starting ST; and follow up at a PCORnet study site. Exclusion criteria include treatment for cancer except mRCC. The primary objective is to determine patterns of change in QOL and symptom burden of pts with mRCC. Early death was defined as death within 12 months post-enrollment. All pts with non-missing FKSI-19 or FKSI-DRS at enrollment were included. Cumulative event probabilities as a function of time were stratified by either < or ≥ median QoL value and plotted as a step function vs time since enrollment to create cumulative event curves. To determine association between QoL at enrollment and early death, separate Cox proportional hazards models (unadjusted and adjusted) for time-to-first-event as a function of baseline QoL were fit. The adjusted model controlled for the following baseline covariates: IMDC risk (favorable vs intermediate or poor), histology (clear cell vs non). Results: As of 1/5/26, 468 pts were enrolled; 425 had non-missing FKSI-19 or FKSI-DRS at enrollment (329 immediate ST, 96 deferred ST [DST]). ST included 129 IO-IO, 108 IO-TKI, 92 Other. DST included 57 Active Surveillance, 22 Local Therapy Planned, 17 Other/Missing. 209 and 190 pts have < median FKSI-19 (score = 59) and FKSI-DRS (score = 29), respectively; 216 and 235 pts have ≥ median FKSI-19 and FKSI-DRS, respectively. Those below the median have lower rates of nephrectomy, worse IMDC risk and KPS, shorter time since RCC diagnosis, and higher rates of bone metastasis (all p<0.05). Cumulative event probabilities for death at 12 months are 21% and 7% in those < and ≥ median baseline FKSI-19, respectively; and 21% and 7% in those < and ≥ median baseline FKSI-DRS, respectively. For FKSI-19, the unadjusted hazard ratio (HR) for early death is 0.55 (95% CI: 0.43 – 0.70; p<0.001) and adjusted HR is 0.65 (0.49 – 0.85; p = 0.002) per 1 SD higher score. For FKSI-DRS, the unadjusted HR for early death is 0.57 (95% CI: 0.46 – 0.71; p<0.001) and adjusted HR is 0.66 (0.51 – 0.86; p =0.002) per 1 SD higher score (higher score indicates better QOL). Conclusions: Across management strategies in our ODYSSEY cohort of systemic therapy naïve mRCC patients, better baseline QOL was associated with better survival. This is to our knowledge the largest reported real world mRCC cohort with prospectively collected PROs. Future analyses will focus on association of outcomes with longitudinal PROs. Clinical trial information: NCT04919122 .

4183 Background: High grade neuroendocrine neoplasms (HG-NENs) are treated with platinum doublets mirroring guidelines for small cell lung cancer (SCLC) but recurrences are common and salvage options are limited in efficacy. Cabozantinib (CABO), an inhibitor of VEGFR, MET, and TAM kinases (TYRO3, AXL, and MER) was approved by the FDA for treatment of well differentiated (WD) NENs of both pancreatic and extrapancreatic origins based on the CABINET study (Chan, NEJM 2025). However, its efficacy in the whole spectrum of HG-NEN patients including poorly differentiated disease (PDD) has not been prospectively explored. Methods: This was a single institution, Phase II study of CABO in patients with HG-NENs who had progressed on at least one prior treatment. Key inclusion/exclusion criteria were NENs of any origin except SCLC, high grade by Ki-67 of > 20% or histology consensus, ECOG of < = 1, appropriate hematological parameters and no history of bleeding or active cardiac disease. CABO was given orally starting at 60 mg po daily in 3-week cycles. The primary endpoint of this study was overall response rate (ORR). Secondary endpoints included progression free survival (PFS) and overall survival (OS). A Simon optimal 2-stage design tested the null hypothesis that the true ORR is < = 1% at the type I error rate of 5%, resulting in projected enrollment of up to 32 total patients. Toxicities were graded according to CTCAE v5.0 and response was evaluated according to RECIST v1. Results: All patients have been accrued, with 4/32 patients still on treatment. Median age at diagnosis was 62 years and male to female ratio was 1.46. Origin of tumor was GI in 68.8% of the cases, the rest being thoracic (6.3%), prostate (6.3%) cervical (6.3%) head/neck (3.1%) and unknown (9.4%). Median Ki-67 was 55%, 11 patients had Ki-67 > 70%. Histology was WD-HG in 40% of patients and PDD in 60%. Two patients had mixed PDD/other components (MiNEN). Median PFS in evaluable patients was 4.01 months [95% CI 1.61 to 9.30] and mOS was 9.89 mo [95% CI 6.73 to 18.96]. Three patients so far (9.3%) have had partial response as best response (PDD-colon, WD-pancreas and PDD-cervical) while 19/32 (59.4%) patients have had stable disease as best response so far (disease control rate 68.8%). Five (15.6%) patients have received more than 10 treatment cycles. The three longest treated patients had WD-pancreas (23 and 17 cycles) and PDD-unknown (12 cycles). One patient had eventual resection of a metastatic site and is currently without evidence of disease. Three patients withdrew from the trial because of toxicities (shortness of breath, GI bleeding, thromboembolism). Twenty-six patients have expired. Conclusions: CABO monotherapy showed efficacy in some HG-NEN patients with not just WD but also the very aggressive PDD histology. No new side effects to the ones previously described for this agent were noted. Supported by Exelixis and a Siteman investment program grant. Clinical trial information: NCT04412629 .

4122 Background: BTCs are rare and aggressive malignancies. The first-line standard of care regimen for advanced BTC (aBTC) is gemcitabine (GEM), cisplatin, and an immune checkpoint inhibitor. Although FOLFOX is a preferred second-line treatment, it is limited by neuropathy. Nal-IRI contains IRI free base in liposome nanoparticles, which shelter IRI from conversion to its active metabolite (SN-38) and increase intratumoral SN-38 compared with IRI alone. The NAPOLI-1 trial of FU/LV/nal-IRI vs. FU/LV in second-line advanced pancreatic adenocarcinoma showed an overall survival (OS) benefit. The NIFTY trial (South Korea) demonstrated median OS (mOS) benefit of FU/LV/nal-IRI over FU/LV in a second-line aBTC population (8.6 vs. 5.3 months [mo], HR 0.68, 95% CI 0.48-0.95); however, the NALIRICC trial (Germany) did not (6.9 vs. 8.2 mo, HR 1.08, 95% CI 0.68-1.72). We sought to characterize second-line FU/LV/nal-IRI efficacy in US patients (pts) with second-line aBTC. Methods: This was a single-arm, open-label, multicenter phase II study of pts with aBTC previously treated with GEM/platinum chemotherapy. Pts received nal-IRI 70 mg/m 2 IV over 90 minutes, LV 400 mg/m 2 IV over 30 minutes every 14 days, and FU 2400 mg/m 2 IV over 46 hours every 14 days. The primary objective was to determine progression-free survival rate at 4 mo (PFS 4mo ) using RECIST v. 1.1 criteria. Median PFS reported for pts receiving second-line 5-FU doublet chemotherapy was 3 mo with a PFS of 30%. FU/LV/nal-IRI was of interest if it could increase the PFS 4mo to 50% or higher. Using a one-sided α of 0.05 and 80% power, 17 of 39 evaluable pts had to be progression-free at 4 mo to detect a difference in PFS 4mo between 30% and 50%. Results: Forty-eight pts were enrolled: median age 64.5 years; 69% women; 27% gallbladder primary; 54% intra- and 19% extrahepatic. Ten pts came off study due to toxicity without disease progression before 4 mo on study and were replaced to evaluate the PFS 4mo endpoint. Of 38 evaluable pts, 18 pts (PFS 4mo 47%) were alive and progression-free at 4 mo, meeting the primary endpoint. With a median follow-up of 7.1 mo, median time to disease progression or study discontinuation due to toxicity for all pts was 2.3 mo (95% CI 1.8-4.6, PFS 4mo 38%), and for pts evaluable for the primary endpoint, median PFS was 3.5 mo (95% CI 1.8-5.5). Five pts (10%) had PFS beyond 12 mo. mOS for all pts was 7.9 mo (95% CI 5.7-11.7). Of 41 evaluable pts, the response rate was 2% (1 PR), and the disease control rate was 54%. Adverse events (AEs) were consistent with known toxicities from the FU/LV/nal-IRI regimen. Conclusions: FU/LV/nal-IRI is an effective second-line regimen for aBTC in US pts. There was significant early toxicity as 21% of pts came off study due to AEs before 4 mo. Correlative studies, including longitudinal analyses of circulating tumor DNA using banked blood specimens, are planned (NCT04005339). Clinical trial information: NCT04005339 .

Estimates of fibromyalgia prevalence vary widely worldwide due to differences in diagnostic criteria, study methodology, and population characteristics. Accurate estimates are essential to inform population-level surveillance, guide resource allocation, and highlight disparities in disease recognition and access to care. The objective of this systematic review and meta-analysis was to determine the global prevalence of fibromyalgia in the general population, and secondarily to identify factors moderating prevalence variation and to assess prevalence in defined at-risk subpopulations. Pooled prevalence estimates with 95% confidence intervals (CIs) were calculated using the inverse variance heterogeneity model with Freeman–Tukey transformation. Subgroup analyses were performed by sex, diagnostic criteria, country, and other study characteristics, and meta-regression assessed moderators including publication year and Human Development Index (HDI). Of 21,645 records identified, 1,728 underwent full-text review and 882 studies were included in the final analysis. Across 38 countries and 30,070,032 individuals (188 studies), the pooled fibromyalgia prevalence in the general population was 1.40% (95% CI, 0.49–2.47), with substantial heterogeneity (I 2 = 100%) and very low certainty of evidence. Among at-risk subpopulations (9,771,123 participants; 768 studies), prevalence was markedly higher in autoimmune and inflammatory conditions, such as chronic urticaria (37.1%), interstitial cystitis (30.8%), systemic sclerosis (20.6%), celiac disease (20.5%), and inflammatory bowel disease (19.9%), and elevated in chronic pain, rheumatologic, and post-infection cohorts. Meta-regression identified publication year (β = 0.008, P < .001) and HDI (β = 0.725, P = .002) as significant moderators of prevalence. In conclusion, fibromyalgia affects 1% to 2% of the global population, with substantially higher rates in at-risk subpopulations. Temporal and socioeconomic trends emphasize diagnostic inconsistency and global disparities in care.

ABSTRACT Interferon-γ (IFNγ) signaling is prominent in inflammatory CNS microenvironments across many neurological disorders, but the neuronal peptides presented on HLA class I under these conditions and their functional consequences for CD8 + T cells remain incompletely defined. Here we combine human iPSC-derived human neural aggregates (HNAs), HLA class I immunoprecipitation coupled to LC-MS/MS immunopeptidomics, and microfluidic co-culture assays to map IFNγ-induced neuronal antigen presentation and test antigen-specific cytotoxicity. IFNγ stimulation induced robust HLA class I expression in HNAs and enabled recovery of a canonical 8-12-mer class I ligandome enriched for 9-mers. Neuron-restricted expression of a synapsin-driven polyepitope cassette yielded presentation of defined neoantigen 9-mer peptides on donor HLA class I molecules and, in the presence of IFNγ, elicited activation of autologous antigen-specific CD8 + T cells and consequent antigen-dependent neurite injury. Across four donors, comparative immunopeptidomics identified large IFNγ-unique neural peptide repertoires distinct from matched fibroblasts and revealed a consistent enrichment of predicted high-affinity binders on HLA-B allotypes. Finally, β2-microglobulin deletion ablated peptide recovery, and neuron-restricted β2-microglobulin reconstitution enabled identification of neuron-derived peptides, including peptides derived from neurofilament light (NEFL) that were shared across donors and presented on multiple HLA allotypes. Together, these data provide an integrated platform for neuronal autoantigen discovery and functional validation and support a model in which IFNγ-driven neuronal HLA class I presentation creates an HLA-B-weighted epitope landscape that can be recognized by autoreactive cytotoxic CD8 + T cells.