Radiation pneumonitis (RP) is a common and clinically significant toxicity of thoracic radiation therapy that can cause pulmonary morbidity and impair quality of life. Although conventional dose-volume histogram-based metrics and normal tissue complication probability models are widely used for RP risk assessment, they inadequately capture the complex spatial, anatomical, and patient-specific factors underlying radiation-induced lung injury. Recent machine learning approaches have improved RP risk prediction by integrating multimodal clinical and imaging information; however, most provide a point risk estimate without quantifying the reliability of individual predictions, limiting their potential clinical utility. We propose a Multimodal Bayesian Diffusion Transformer (MM-DiT) framework that jointly estimates RP risk and characterizes the sources of predictive uncertainty. MM-DiT integrates planning computed tomography (CT) images and three-dimensional radiation dose distributions through self-supervised multimodal pre-training, reducing reliance on limited and potentially noisy toxicity labels. The resulting representations are further refined using a latent diffusion transformer and transferred to a Bayesian prediction framework for probabilistic RP risk estimation. A learnable label-noise model is incorporated to explicitly account for uncertainty arising from imperfect toxicity annotations. Therefore, it provides individualized RP risk estimates with complementary measures of aleatoric, epistemic, and label uncertainty, enabling assessment of prediction reliability at the individual-patient level. We evaluated MM-DiT in two independent cohorts using complementary assessments of predictive discrimination, calibration, and uncertainty. The results demonstrate its potential to provide accurate RP risk estimates while quantifying clinically relevant sources of predictive uncertainty.

Neoantigen-targeted immunotherapies hold promise for cancer treatment, but current personalized approaches are time-consuming and costly. Here, we identify neoantigens encoded by Ptprs and Igf2r that are shared across murine mismatch repair-deficient colorectal and breast tumors and unexpectedly conserved in human colorectal, endometrial, gastric, and prostate cancers. These neoantigens elicit spontaneous, organ-spanning CD8+ T cell-mediated memory responses that are enhanced by immune checkpoint blockade. Vaccination with mRNA/lipid nanoparticles encoding these conserved neoantigens suppresses tumor growth across prophylactic and therapeutic models, including checkpoint-resistant orthotopic tumors. Tumor rejection is accompanied by antigen spreading, abscopal effects, and infiltration by clonally diverse T cells, dendritic cells, and MHC I/II+ macrophages producing CXCL9/10, CCL5/8, and TNF. Tumor cells also show activation of innate and adaptive pathways, including MHC and ISGs overexpression. Our results uncover a conserved anti-tumor immune mechanism and support the development of off-the-shelf neoantigen vaccines across tissues and species. Guillaume Mestrallet, Ross Ward, Matthew Brown, Jesse Boumelha, Frederika Rentzeperis, Natalie Vaninov, Miriam Saffern, Ezekiel Olumuyide, Prerna Suri, Sreekumar Balan, Leandra Velazquez, Aparna Ananthanarayanan, Zhihong Chen, Aimee Lucas, Miriam Merad, Cansu Cime-Bozkus, Nicolas Vabret, Robert Samstein, Nina Bhardwaj. Targeting neoantigens conserved across organs and species overcomes tumor immune escape [abstract]. In: Proceedings of the AACR Immuno-Oncology Conference (AACR IO): Discovery and Innovation in Cancer Immunology: Revolutionizing Treatment through Immunotherapy; 2026 Feb 18-21; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Immunol Res 2026;14(2 Suppl):Abstract nr C070.

BACKGROUND:Knowledge-based planning (KBP) has improved the quality and efficiency of radiotherapy treatment planning. However, its broader clinical adoption remains limited because effective deployment often requires institution-specific model training and tuning. Publicly available KBP models provide a convenient starting point but may not consistently meet local clinical objectives across institutions. PURPOSE:We developed and evaluated the Planning Copilot, a large language model (LLM)-guided plan refinement framework designed to operate as a model-agnostic post-processing layer for KBP. METHODS:The Planning Copilot is a closed-loop, multi-agent system that iteratively refines KBP-generated plans through structured dosimetric feedback and the selection of clinically validated optimization actions within a treatment planning system. For each case, an initial step-and-shoot IMRT plan was generated with each of three RapidPlan models, including a publicly available model and two institutional models with different optimization constraints. To assess whether the refinement depends on KBP, we additionally evaluated PlanningCopilot starting from a non-KBP fixed objective template applied identically to all cases. The PlanningCopilot was applied without model-specific tuning to 62 retrospective locally advanced NSCLC cases. Clinical goal achievement rates and clinically relevant dose-volume metrics were compared between the initial KBP plans and the refined plans. RESULTS:Across all three KBP models, the PlanningCopilot substantially improved plan quality. Clinical goal achievement increased from 79% to 98% for the UCSD model, from 73% to 97% for Institutional T1, and from 69% to 98% for Institutional T2. Starting from the non-KBP fixed template, the achievement rate increased from 68% to 97%, comparable to the KBP initializations. Significant reductions were observed in key lung dose metrics, including lung Dmean across all models and lung V20 in the Institutional T1 model and template initializations, while target coverage and doses to critical structures were maintained. Notably, the KBP model that prioritized OAR sparing, which exhibited the lowest initial pass rate, showed the highest rescue rate after refinement. CONCLUSIONS:An LLM-guided refinement layer can improve the success rate and portability of KBP across heterogeneous models without retraining the underlying KBP system. This approach provides a practical strategy to enhance the reliability of KBP and supports the use of off-the-shelf models through automated, model-agnostic post-processing.

BACKGROUND:Consistently generating clinically acceptable plans without human intervention remains a challenge in radiotherapy. Rule-based automation provides deterministic execution, and knowledge-based planning (KBP) provides statistical dose estimation, but both often require manual refinement. Large language models (LLMs) offer clinical reasoning capability, but effective autonomous planning also requires a mechanism to execute complex planning actions within the treatment planning system (TPS). PURPOSE:To develop and evaluate PlanningCopilot, an agentic system that utilizes the reasoning capability of LLM and a validated Eclipse Scripting API (ESAPI) optimization module integrating KBP initialization ("PlanAct") to autonomously generate treatment plans. This study evaluates the system's ability to produce clinically acceptable plans for locally advanced non-small cell lung cancer (LA-NSCLC) and assesses its potential to refine performance by self-learning. METHODS:PlanningCopilot was implemented as a multi-agent framework linked to the TPS through PlanAct API. It comprises four specialized GPT-4.1 agents that iteratively interact with the TPS: (1) an Evaluator agent that accesses the plan and generates plan quality reports, (2) a Supervisor agent that validates these reports before passing them to a Planner agent, (3) the Planner agent that executes initialization and optimization tasks through PlanAct API and planning guidelines, and (4) an optional Learner agent that synthesizes optimization history into Planner-facing prompt addendums. We retrospectively analyzed 62 patients with conventionally fractionated LA-NSCLC and compared original clinical plans with autonomous plans with and without the Learner agent. Measurement-based patient-specific quality assurance (PSQA) was performed on the first 21 autonomous IMRT plans in planning order. RESULTS:All autonomous plans met clinical dosimetric requirements, including those not achieved in the clinical plans and KBP (RapidPlan) plans. Paired Wilcoxon signed-rank tests showed no significant differences between autonomous and clinical plans for Lungs Dmean (p = 0.371), Lungs V20Gy (p = 0.449), Lungs V5Gy (p = 0.309), Heart D50% (p = 0.175), Esophagus Dmean (p = 0.750), Spinal Cord D0.03cc (p = 0.422), and Plan D0.03cc (p = 0.941). Furthermore, autonomous plans achieved significantly lower Esophagus D0.03cc (p = 0.027). Compared with RapidPlan initialization, PlanningCopilot improved multiple dosimetric endpoints, including Lungs Dmean (p < 0.001), Lungs V20Gy (p < 0.001), Lungs V5Gy (p = 0.004), Heart D50% (p = 0.037), and Plan D0.03cc (p < 0.001), with the cost of higher Esophagus Dmean (p < 0.001) and Esophagus D0.03cc (p < 0.001). In a subset of 18 cases requiring at least two iterations, applying Learner-derived knowledge reduced required iterations by an average of 11.8% while maintaining comparable plan quality (p > 0.05). All 21 autonomous IMRT plans passed measurement-based PSQA. CONCLUSION:PlanningCopilot enables autonomous generation of clinically acceptable and deliverable treatment plans for LA-NSCLC. It consistently satisfies clinical dosimetric requirements across varying anatomical complexities and improves optimization efficiency through self-learning from prior optimization history.

BACKGROUND:Knowledge-based planning (KBP) is a data-driven approach that utilizes the knowledge from previous high-quality treatment plans to predict dose-volume histogram (DVH) parameters for organs-at-risk (OARs) in new cases. Research has demonstrated that KBP enhances plan quality, minimizes inter-patient and inter-institution variability, and significantly boosts time efficiency. However, current state-of-the-art KBP approaches only generate one set of planning goals for one point on the Pareto optimal surface without considering potential planning trade-offs. PURPOSE:The objective of this study is to develop a KBP trade-off prediction model that can effectively assist clinical decision-making during the treatment planning process for patients with locally advanced non-small cell lung cancer (NSCLC). METHODS:We created 13 volumetric-modulated arc therapy (VMAT) plan variations for each patient in our dataset (n = 53), consisting of one balanced plan and 12 alternative plans with trade-off considerations. These trade-off plans incorporated three levels (0-2) of sparing priority for each OAR, including the esophagus, lungs, heart, and spinal cord. The first three principal components (PCs) of each OAR-specific DVH were used as target variables, while 26 anatomical features served as predictors. The patients were randomly divided into a training set (80%) and a test set (20%). A forward feature selection process identified the top five anatomical features, which were then used to train a random forest multi-output regression model to predict the first three DVH PCs for each of the 13 plan-OAR variations. We compared the performance of our trade-off prediction model with that of a balanced model trained on the balanced plan without any trade-off considerations. The evaluation metrics included root-mean-square error (RMSE) and mean absolute error (MAE) for key dose-volume metrics of the DVH curves. RESULTS:The trade-off prediction model significantly outperformed the balanced model in terms of average RMSE (5.32 vs. 27.3) compared to the planned DVHs for all 13 treatment plans. The trade-off model also achieved a lower MAE for all the clinical dose-volume metrics, including spinal cord Dmax (12.5 vs. 15.5 Gy, p < 0.01), esophagus Dmax (1.7 vs. 2.7 Gy, p < 0.01), left lung V20Gy (7.8% vs. 27.5%, p < 0.01), right lung V20Gy (7.4% vs. 27.8%, p < 0.01), and heart V30Gy (10.8% vs. 20.9%, p < 0.01). CONCLUSION:Our proposed KBP trade-off model reliably predicts plan tradeoff variations for the treatment of NSCLC patients. Incorporating this model into the pre-planning process may serve as a decision support tool for physicians and planners by providing feasible trade-off estimations, thereby improving the efficiency of the treatment planning workflow.

Basic and translational research in lung cancer is a rapidly evolving field with a transformational impact on early detection, diagnosis, therapeutic development, and personalization of care. Recent advances have greatly increased our understanding of the molecular genomics, proteomics, pathogenesis, and cellular biology of this deadly malignancy. The International Association for the Study of Lung Cancer (IASLC) recently formed a Basic and Translational Science (BaTS) Committee to further enhance the scientific leadership of IASLC in thoracic cancer research. This review by members of the committee highlights the breadth of current research in NSCLC, with a focus on molecular risk factors and processes in tumorigenesis, heterogeneity, phenotypic plasticity, metabolic reprogramming, immunobiology, the immune microenvironment, and microbiome. This review also identifies future research areas that may lead to further improvement in survival outcomes and curative therapies especially for patients with advanced NSCLC.

Purpose This study aims to evaluate the impact of varying definitions of normal lung volume on the prediction of radiation pneumonitis (RP) risk in patients with locally advanced non-small cell lung cancer undergoing radiation therapy. Methods and Materials Dosimetric variables V20, V5, and mean lung dose (MLD) were extracted from the treatment plans of 442 patients enrolled in the NRG Oncology Radiation Therapy Oncology Group 0617 trial. Three different definitions of lung volume were evaluated: total lung excluding gross tumor target, total lung excluding clinical target volume, and total lung excluding planning target volume (TL-PTV). Patients were grouped as “no-RP2” (N = 377, grade ≤1 RP) and “RP2” (N = 65, grade ≥2 RP). Statistical analyses were performed to assess the effect of lung volume definition on RP2 prediction. Three supervised machine learning models—logistic regression, k-nearest neighbor (kNN), and eXtreme Gradient Boosting—were used to evaluate predictive performance. Model performance was quantified using the area under the receiver operating characteristic curve, and statistical significance was tested via a bootstrap analysis. Shapley Additive Explanations (SHAP) were applied to interpret feature contributions to model predictions. Results Statistical analyses showed that V20 and MLD were significantly associated with RP2, while differences among the lung volume definitions were not statistically significant. Both k-nearest neighbor and eXtreme Gradient Boosting classifiers consistently yielded higher area under the receiver operating characteristic curve values for the TL-PTV definition compared to the other definitions, a finding supported by bootstrap analysis. SHAP analysis further indicated that V20 and MLD were the most influential predictors of RP2. Conclusions In line with previous studies, both statistical analysis and SHAP interpretation confirmed that V20 and MLD were associated with RP2. The machine learning models indicated that defining normal lung volume as TL-PTV yielded the highest predictive performance for RP2 risk. Further validation using external data sets are warranted to confirm these findings.

Colorectal cancer (CRC) accounts for 10% of all cancer cases and is the second leading cause of cancer-related deaths worldwide. Immunotherapies have significantly advanced over the past decades, marking a major breakthrough in cancer treatment. While anti-PD-1 therapy is utilized in both local and advanced disease, up to 50% of mismatch repair deficient (MMRd) and most mismatch repair proficient (MMRp) CRC fail to respond. Using orthotopic animal and patient-derived models of CRC, along with single cell and spatial analyses, we determined that colocalization and interactions between MHC + C1Q + CXCL9 + macrophages and TCF + PRF1 + T cell subsets are associated with control of tumor growth during anti-PD-1 treatment. In contrast, resistance is associated with upregulation of TIM3, LAG3, TIGIT, and PD-1 expression on T cells, and tumor infiltration by immunosuppressive TREM2 + macrophages and monocytes in T cell excluded zones. A novel combinatorial checkpoint blockade targeting TREM2, LAG3, CTLA4, and PD-1 achieves up to 100% tumor clearance in MMRd CRC and > 70% in MMRp CRC models, compared to 0% with anti-PD-1 monotherapy. This approach induces durable anti-tumor immune memory, mediated by coordinated interactions among MHC + macrophages, CD4+/CD8 + T cells, and TCF + T cells. It also reduced infiltration by immunosuppressive myeloid cells and T cell exhaustion. Together this study identifies key T cell and macrophage subsets mediating the efficacy of immunotherapy in overcoming immune escape in both MMRd and MMRp CRC settings.

Mismatch repair deficiency (MMRd), either due to inherited or somatic mutation, is prevalent in colorectal cancer (CRC) and other cancers. While anti-PD-1 therapy is utilized in both local and advanced disease, up to 50% of MMRd CRC fail to respond. Using animal and human models of MMRd, we determined that interactions between MHC+ C1Q+ CXCL9+ macrophages and TCF+ BHLHE40+ PRF1+ T cell subsets are associated with control of MMRd tumor growth, during anti-PD-1 treatment. In contrast, resistance is associated with upregulation of TIM3, LAG3, TIGIT, and PD-1 expression on T cells, and infiltration of the tumor with immunosuppressive TREM2+ macrophages and monocytes. By combining anti-PD-1 with anti-LAG3/CTLA4/TREM2, up to 100% tumor eradication was achieved in MMRd CRC and remarkably, in >70% in MMRp CRC. This study identifies key T cell and macrophage subsets mediating the efficacy of immunotherapy in overcoming immune escape in both MMRd and MMRp CRC settings.

Pathogenic germline variants in mismatch repair (MMR) genes, or Lynch syndrome (LS), increases patients’ risk of developing colorectal cancer (CRC). We previously identified shared, frameshift (fs)-neoantigens in MMR deficient (MMRd) cancers. We hypothesized that T cell surveillance occurs early in CRC development in LS but T cell activity is abrogated by immune regulatory programs during tumor progression. Blood and normal, precancerous (adenoma), and CRC tissues were analyzed from a cohort of 92 LS patients. We leveraged a custom neoantigen-discovery pipeline, functional antigen recognition assays, T cell receptor sequencing, and single cell and spatial transcriptomics. This identified for the first time T cells recognizing shared fs-neoantigens expressed in normal mucosa, adenomas, and tumors of LS patients. Immune editing was evidenced by i) distinct neoantigen repertoires in precancerous tissues compared to tumors and ii) recurrence of low affinity (but not high affinity) neoantigens in later lesions captured from the same patients. Transcriptomic and ex vivo functional analysis revealed that T cells in tumors, relative to precancerous tissues, had reduced functional capacity associated with an infiltration of immunoregulatory myeloid cells (tumor-associated neutrophils and TREM1+ macrophages) which can be therapeutically targeted in future clinical studies. Finally, these shared, precancerous fs-neoantigens can serve as vaccine targets to prevent MMRd cancer in LS. Supported by NIH UG3CA290517 and the Parker Institute for Cancer Immunotherapy Tumor Immunology: Cellular Responses and Tumor Microevironment (TIME)

A preprint by Zhijie et al. identifies somatic mutations in TYK2 that are enriched in tumour-infiltrating lymphocytes and associated with an increased antitumour response.

Studies by our lab have demonstrated that DNA Damage Repair (DDR) pathways play a key role in determining response to ICB. To better characterize the effect of these alternations on the tumor immune microenvironment (TME) and define key determinants of ICB response, we created isogenic knockouts of BRCA2 and BRCA1 in a murine 4T1 metastatic TNBC background to assess how these tumor-intrinsic programs influence the TME and poise tumors for immunotherapy response. Differential immune landscapes identified via scRNAseq in Brca2-mutant tumors at baseline showed key differences in the myeloid compartment and interferon-stimulated gene (ISG) expression in tumor-infiltrating myeloid cells from Brca2 mutant tumors. Bulk RNAseq analysis showed increased production of T cell trafficking chemokines that could poise the tumor for response to ICB. Interferon reporter assay results suggested tumor-intrinsic cGAS drives trans-activation of myeloid STING and IFNb1. Further assessment of tumor cell lines via cellular fractionation experiments identified the presence of both DNA and R-loops in the cytoplasm of BRCA2-mutant cell lines that serve to active cGAS/STING. In vivo experiments with tumor intrinsic knockouts showed that tumor STING signaling and cytokine production were dispensable for ICB response while depletion of monocytes via CSF1R blockade showed a complete reversal of ICB response, underlining the essential role of monocytes in poising the tumor for ICB response. Supported by NIH Director’s Early Independence Award, the Parker Institute for Cancer Immunotherapy, Burrough’s Welcome Fund, and Mount Sinai’s Cancer Biology T32 Training Grant. Tumor Immunology: Cellular Responses and Tumor Microevironment (TIME)

Predicting whether a patient with cancer will benefit from immune checkpoint inhibitors (ICIs) without resorting to advanced genomic or immunologic assays is an important clinical need. To address this, we developed and evaluated SCORPIO, a machine learning system that utilizes routine blood tests (complete blood count and comprehensive metabolic profile) alongside clinical characteristics from 9,745 ICI-treated patients across 21 cancer types. SCORPIO was trained on data from 1,628 patients across 17 cancer types from Memorial Sloan Kettering Cancer Center. In two internal test sets comprising 2,511 patients across 19 cancer types, SCORPIO achieved median time-dependent area under the receiver operating characteristic curve (AUC(t)) values of 0.763 and 0.759 for predicting overall survival at 6, 12, 18, 24 and 30 months, outperforming tumor mutational burden (TMB), which showed median AUC(t) values of 0.503 and 0.543. Additionally, SCORPIO demonstrated superior predictive performance for predicting clinical benefit (tumor response or prolonged stability), with AUC values of 0.714 and 0.641, compared to TMB (AUC = 0.546 and 0.573). External validation was performed using 10 global phase 3 trials (4,447 patients across 6 cancer types) and a real-world cohort from the Mount Sinai Health System (1,159 patients across 18 cancer types). In these external cohorts, SCORPIO maintained robust performance in predicting ICI outcomes, surpassing programmed death-ligand 1 immunostaining. These findings underscore SCORPIO’s reliability and adaptability, highlighting its potential to predict patient outcomes with ICI therapy across diverse cancer types and healthcare settings. Using multiple datasets from real-world evidence and completed trials, a machine learning model using routine blood and clinical data is shown to be predictive of patient response to immune checkpoint inhibitor therapy, across cancer types and outperforming standard biomarkers.

ADAR1 is an RNA editing enzyme which prevents autoimmunity by blocking interferon responses triggered by cytosolic RNA sensors, and is a potential target in immuno-oncology. However, predictive biomarkers for ADAR1 inhibition are lacking. Using multiple in vitro and in vivo systems, we show that BRCA1/2 and ADAR1 are synthetically lethal, and that ADAR1 activity is upregulated in BRCA1/2-mutant cancers. ADAR1 depletion in BRCA1-mutant cells causes an increase in R-loops and consequently, an upregulation of cytosolic nucleic acid sensing pattern recognition receptors (PRR), events which are associated with a tumor cell-autonomous type I interferon and integrated stress response. This ultimately causes autocrine interferon poisoning. Consistent with a key role of R-loops in this process, exogenous RNase H1 expression reverses the synthetic lethality. Pharmacological suppression of cell-autonomous interferon responses or transcriptional silencing of cytosolic nucleic acid sensing PRR are also sufficient to abrogate ADAR1 dependency in BRCA1-mutant cells, in line with autocrine interferon poisoning playing a central part in this synthetic lethality. Our findings provide a preclinical rationale for assessing ADAR1-targeting agents in BRCA1/2-mutant cancers, and introduces a conceptually novel approach to synthetic lethal treatments, which exploits tumor cell-intrinsic cytosolic immunity as a targetable vulnerability of cancer cells.

(Wilson* et al Science 2024 & unpublished work ) Cancer risk is modulated by a complex network of inherited mutations, DNA replication errors, and environmental exposures. Yet, the impact of genetic variation in the immunosurveillance of nascent and established malignancies remains uncertain. Using population-scale data from the UK Biobank and FinnGen, we uncover a striking association between HLA allelic-associated peptidome diversity and reduced lung cancer risk in smokers. Fine-mapping reveals that amino acid heterozygosity in the HLA-II peptide-binding groove significantly contributes to this protective effect. Single-cell analyses further demonstrate that smoking induces proinflammatory lung macrophages and HLA-II+ epithelial cells, emphasizing a dynamic immune-environment interaction. Furthermore, we identify widespread loss of HLA-II heterozygosity (LOH) in lung cancer, favoring alleles with expanded neopeptide repertoires, alongside distinct LOH patterns for HLA-I and HLA-II across cancer types. Building on these insights, we introduce a novel biophysically informed embedding space, developed with state-of-the-art large language models, to represent HLA allelic similarity as a continuous variable. This framework enables precise quantification of risk across diverse cancers. Together, our findings position genetic variation in immunosurveillance as a pivotal determinant of cancer risk, offering new avenues for predictive modeling and therapeutic innovation. US NIH grant R01 CA283469 Alexander and Alexandrine Sinsheimer Foundation Tumor Immunology: Checkpoints, Prevention, and Treatment (TIPT)

BACKGROUND:Manual intensity-modulated radiotherapy (IMRT) planning for locally advanced non-small cell lung cancer (LA-NSCLC) is labor-intensive and time-consuming. Knowledge-based planning (e.g., RapidPlan) improves consistency but commonly falls short in fully meeting clinical objectives in LA-NSCLC cases, requiring iterative manual adjustments. PURPOSE:To develop and validate PlanAct, an Eclipse Scripting API (ESAPI)-based module for optimizing automated IMRT planning in LA-NSCLC, and to compare its performance against clinical and RapidPlan-generated plans across a retrospective patient cohort. METHODS:PlanAct was developed with modular functions to automate key tasks in IMRT plan generation and optimization. PlanAct was manually executed on 56 anonymized retrospective LA-NSCLC cases using a standardized nine-beam geometry. Plans were normalized to ensure 95% planning target volume (PTV) coverage. The PlanAct-optimized plans were evaluated against RapidPlan-generated plans and clinically approved plans using institutional plan quality metrics, including dose-volume constraints for the esophagus, spinal cord, lungs, heart, larynx, and PTV. Statistical comparisons were performed to assess differences in plan quality and unmet dosimetric requirements. RESULTS:PlanAct-optimized plans demonstrated significant improvement in plan quality compared to RapidPlan, with fewer unmet clinical requirements and better organ-at-risk sparing, particularly for the lungs (p < 0.001 for V20 and Dmean). Only one PlanAct-optimized plan failed to meet all dose constraints (in this case, lungs Dmean) due to a large PTV volume, compared to 18 RapidPlan and 10 clinical plans. Even in anatomically challenging cases, PlanAct produced more favorable dose distributions, with superior hotspot control. CONCLUSIONS:PlanAct is an effective tool to optimize automated IMRT planning in LA-NSCLC. It produced plans comparable to or better than clinical plans, even in challenging cases. Its modular architecture makes it promising for integration into future fully autonomous, patient-specific radiotherapy treatment planning systems.

Chronic inflammation can contribute to growth and immune evasion of solid tumors, with multiple cytokines and chemokines implicated in this process. We previously showed that disrupting the Th2 immune response systemically by blocking IL-4 signaling can activate dendritic cells and T effector cells to generate a robust immune response against tumor antigens in pre-clinical lung cancer models (Maier, B., et al., A conserved dendritic-cell regulatory program limits anti-tumour immunity. Nature, 2020.580[7802]:257-262). Further analysis of our pre-clinical lung cancer model has indicated that IL-1 signaling contributes to pathogenic emergency myelopoiesis and progression of lung cancer. Dual disruption of IL-1alpha and IL-1beta using anakinra inhibits tumor growth. In humans, we have demonstrated that dupilumab, an IL-4 receptor alpha antagonist, can rescue response to PD-(L)1 blockade in a subset of patients. Based on our pre-clinical data, we have now initiated a Phase 1b/2 trial of dupilumab and anakinra combined with PD-(L)1 blockade in patients with NSCLC who have progressed on previous PD-(L)1 inhibitors. (Park M.D., et al., Hematopoietic aging promotes cancer by fueling IL-1alpha-driven emergency myelopoiesis. Science 386, eadn0327(2024). DOI:10.1126/science.adn0327) This is a Phase 1b/2 trial in patients with relapsed/refractory NSCLC who have previously received PD-(L)1 blockade are eligible for enrollment. Patients with progressive disease on PD-(L)1 agents continue PD-(L)1 targeted therapy, and three doses of dupilumab are added, administered every three weeks, as was performed in the study we recently completed. In this new trial cohort, alongside continued PD-(L)1 blockade and q3w dupilumab, daily injections of anakinra are administered for the first four weeks. Patients undergo pre- and on-treatment biopsies and blood collection throughout, with cryopreservation of PBMCs and plasma at each collection timepoint. The primary endpoint of Phase 1b is the occurrence of dose limiting toxicities, while the primary endpoint of phase 2 is overall response rate. Secondary objectives include best overall response rate, progression free survival, overall survival, and duration of response. Exploratory objectives include characterization of the immunodynamic changes that occur in patients’ tumor microenvironment and systemic immune milieu. Jacob A. Lowy, Matthew D. Park, Nicholas C. Rohs, Pamela Vaiskauskas, Fionnuala Crowley, Nicholas Venturini, Stephanie Chang, Nelson LaMarche, Clotilde Hennequin, Jessica Le Berichel, Pauline Hamon, Meriem Belabed, Nader Yatim, Jesse Boumelha, Alexis Saffon, Frederika Rentzeperis, Brian Y. Soong, Leanna Troncoso, Laszlo Halasz, Christina Noel, Theodore Chin, Earnest P. Chen, Amanda Reid, Maria Casanova-Acebes, Matthew Su, Ashley Reid, Shira Wood-Isenberg, Darwin D’souza, Travis Dawson, Kai Nie, Zhihong Chen, Jorge E. Gomez, Seunghee Kim-Schulze, Filip K. Swirski, Nicolas Vabret, Udit Chaddha, Dan Feng, Mary B. Beasley, David F. Yankelevitz, Deborah B. Doroshow, Rajwanth Veluswamy, Robert Samstein, Brian D. Brown, Fred R. Hirsch, Sacha Gnjatic, Miriam Merad, Thomas U. Marron. A phase 1b/2 trial of dupilumab given in conjunction with PD-1 or PD-L1 blockade and anakinra in the treatment of relapsed/refractory metastatic NSCLC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_2):Abstract nr CT106.