Background and Purpose: Manual segmentation of gynecologic organs and cervical tumors for radiotherapy planning is time-consuming and variable. Automated segmentation on routine T2-weighted magnetic resonance imaging (MRI) remains limited. The aim of this study was to evaluate a lightweight deep learning model for automated segmentation of gynecologic organs and cervical tumors on T2-weighted MRI.Materials and Methods: This work applied a two-stage lightweight deep learning model (PocketNet) to segment the cervix, vagina, uterus, and tumor(s) on T2-weighted MRI in 102 patients with cervical cancer undergoing definitive radiotherapy. Model performance was assessed using the Dice-Sorensen coefficient (DSC) and 95th percentile Hausdorff distance (Haus95) on internal data and validated on an external dataset. A full nnU-Net model trained on the same internal dataset served as a benchmark for segmentation accuracy and computational efficiency.Results: On the institutional dataset, PocketNet achieved mean DSC values exceeding 70% for tumor segmentation and 80% for organ segmentation. External validation on The Cancer Imaging Archive (TCIA) Cervical Cancer Tumor Heterogeneity (CCTH) collection demonstrated the model's robustness, achieving DSC scores of 67.3% for tumor segmentation and 80.8% for organ segmentation. Compared to the PocketNet architecture, nnUNet achieved similar accuracy but required approximately twice the training time, more than 35 times as many parameters, and 40 times more memory for model storage.Conclusion: The PocketNet architecture provides reliable automated segmentation of gynecologic organs and cervical tumors on T2-weighted MRI, with performance comparable to a full-sized nnUNet while requiring substantially less memory and training time, supporting its potential integration into time-sensitive radiotherapy workflows.

Spatial transcriptomics (ST) provides unprecedented insights into gene expression patterns while retaining spatial context, making it a valuable tool for understanding complex tissue architectures, such as those found in cancers. Seurat, by far the most popular tool for analyzing ST data, uses the Wilcoxon rank-sum test by default for differential expression analysis. However, as a nonparametric method that disregards spatial correlations, the Wilcoxon test can lead to inflated false positive rates and misleading findings. This limitation highlights the need for a more robust statistical approach that effectively incorporates spatial correlations. To this end, we propose a Generalized Estimating Equations (GEE) framework as a robust solution for differential gene expression analysis in ST. We conducted a comprehensive comparison of the GEE-based tests with existing methods, including the Wilcoxon rank-sum test and z-test. By appropriately accounting for spatial correlations, extensive simulations showed that the GEE test with robust standard error, referred to as the Independent GEE, demonstrated superior Type I error control and comparable power relative to other methods. Applications to ST datasets from breast and prostate cancer showed poor calibration of the p-values and potential false positive findings from the Wilcoxon rank-sum test. Our comparative study based on simulations and real data applications suggests that the Independent GEE test is well-suited for ST data, offering more accurate identification of biologically relevant gene expression changes and complementing the Wilcoxon rank-sum test. We have implemented the proposed method in R package "SpatialGEE", available on GitHub.

BACKGROUND:PD-L1 expression and tumor mutational burden (TMB) are biomarkers for immune checkpoint inhibitor (ICI) therapy in non-small cell lung cancer (NSCLC); however, patients harboring oncogenic alterations have limited benefit from ICIs. The impact of oncogenic alterations on TMB and PD-L1 tumor proportion score in lung cytology specimens is poorly understood. Herein, the association between oncogenic alterations, TMB, and PD-L1 in NSCLC cytology specimens is explored. METHODS:Next-generation sequencing results from 312 NSCLC cytology specimens were retrospectively reviewed that interrogate 610 genes and select immuno-oncology signatures. TMB and PD-L1 immunohistochemical expression across oncogenic alterations were analyzed to explore associations. RESULTS:Of the 312 cases evaluated, 192 harbored NSCLC-specific oncogenic alterations. Relative to EGFR-mutated tumors, TMB was significantly higher in KRAS (padj = 2.7 × 10-4), ERBB2 (padj = .023), and BRAF (padj = .023) -mutated tumors but lower in ALK-rearranged tumors (padj = .005). Significantly higher PD-L1 expression was seen in tumors with KRAS (padj = .002) and MET exon 14 (padj = 1.06 × 10-4) when compared to EGFR-mutated tumors. Strong positive correlations between TMB and PD-L1 were observed in ERBB2-, KRAS-, and BRAF-mutated tumors when evaluated as continuous variables. TP53 mutations further enhanced immunogenicity when co-occurring with KRAS, ERBB2, or BRAF mutations but this effect was not observed in EGFR-mutated tumors. CONCLUSIONS:These findings demonstrate distinct TMB and PD-L1 profiles that may identify patients who will benefit from ICI therapy. Cytology specimens provide adequate material for biomarker testing, which underscores their value in guiding immunotherapy decisions.

611 Background: Triple-negative breast cancer (TNBC) is an aggressive molecular subtype that accounts for approximately 15%-20% of all breast cancer diagnoses. We aimed to investigate the utility of presurgical DCE breast MRI as a predictive marker for pathologic complete response (pCR) after neoadjuvant treatment (NAT) in TNBC patients and to compare the predictive value of early versus delayed enhancement for residual tumor detection. Methods: A total of 308 Stage I–III TNBC patients who underwent preoperative DCE-MRI after completion of NAT followed by surgery were enrolled in an IRB-approved prospective clinical trial (NCT02276433). Tumor size was measured using three-dimensional measurements of the index lesion during both the early (1 min) and delayed (6 min) phases of DCE-MRI. Treatment response at surgery (pCR vs. non-pCR) and the pathologic size of residual disease were documented. Correlation between pCR and residual enhancement on DCE-MRI was assessed using McNamar’s test. Spearman’s rank correlation coefficient was used to assess concordance between the longest diameter on MRI and pathology. Differences between longest diameter on DCE-MRI and pathology were compared using the Wilcoxon signed-rank test. Results: Among the 308 TNBC patients, 47% (145/308) achieved pCR following treatment. Residual disease detection on the early phase of DCE-MRI demonstrated higher sensitivity for predicting pCR compared to the delayed phase (79% vs. 69%, p < 0.001); however, it had lower specificity (78% vs. 84%, p = 0.008). Absence of enhancement in both early and delayed phase DCE-MRI predicted pCR with positive predictive values (PPV) of 80% and 83%, respectively. Residual enhancement in both phases predicted non-pCR with negative predictive values (NPV) of 77% and 71%, respectively. Both early and delayed DCE-MRI phases demonstrated a similar moderate positive correlation with pathology (r = 0.64 vs. 0.62). There was no significant difference between the longest diameter measured on early phase DCE-MRI and pathology (p = 0.706), whereas a significant difference was observed for the delayed phase (p < 0.001), which over estimated residual disease. Conclusions: Presurgical DCE-MRI demonstrated strong performance in predicting pCR among TNBC patients following NAT. The early and delayed phases of DCE-MRI may each capture different aspects of tumor characteristics, potentially providing complementary information for prediction. Clinical trial information: NCT02276433 .

Supplementary Figure 5. ADAR1 restricts activation of interferon signaling pathway.

Abstract Introduction: Understanding the mechanisms driving evolution of the mucosal field effects to invasive cancer is not possible unless they are analyzed in the context of their geographic distribution in the entire organ. Bladder cancer is an ideal model disease for such studies as it effects an anatomically simple organ permitting the multi-platform analysis of the affected mucosa on the scale of whole organ. Methods: We performed comprehensive multi-platform analyses on nine cystectomies with invasive bladder cancer comprising of 433 mucosal tissue samples analyzed by bulk whole-exome and mRNA sequencing, genome-wide copy number variation and methylation profiling complemented with proteomics, metabolomic and single cell sequencing spatial mapping. The data from multi-platform profiling were geographically annotated to microscopically normal urothelium (NU) and low-grade intraurothelial neoplasia (LGIN; n=243), high-grade intraurothelial neoplasia (HGIN; n=90), and invasive urothelial carcinoma (UC; n=100). Results: We identified ∼16000 non-silent mutations per cystectomy. In two maps the hypermutator phenotypes with ∼48000 and ∼57000 mutations were detected. The mutational analysis identified three types of mutations based on the variant allele frequency and geographic distribution referred to as α, β, and ϒ. Time modeling by a parsimonious time-continuous Markov model incorporating cell migration (immigration) and growth (branching) revealed that bladder carcinogenesis takes approximately three decades and can be divided into dormant and progressive phases. Low selection α mutations were private and most frequent. They continuously developed over 30 years primarily in the dormant phase of carcinogenesis. β mutations clonally expanded regionally and signified the advent of progressive phase of carcinogenesis which lasted five years. ϒ mutations were the ultimate drivers of the progressive phase and emerged 2-3 years before the final progression to invasive bladder cancer. The mutational landscape developed on the background of severely dysregulated urothelial differentiation and increased immune infiltration with T-cell exhaustion. The proteomic and metabolomic changes involved a wide-spread disorganization of glycolipid energy metabolism with downregulation of mitochondrial oxidative phosphorylation. Conclusion: Dysregulated mitochondrial energy metabolism converging on anerobic glycolysis and oxidative phosphorylation with Warburg phenotype emerged as the leading mechanism driving the progression of mucosal field effects to invasive cancer. Citation Format: Bogdan A. Czerniak, Sangkyou Lee, Khanh Ngoc Dinh, Huiqin Chen, Yishan Wang, Jiansong Chen, June Goo Lee, Sung Yun Jung, Nagireddy Putluri, Neema Navai, David McConkey, Charles Chuanhai Guo, Peng Wei, Marek Kimmel. Modeling of bladder cancer evolution from field effects by multi-platform spatial mapping on the whole-organ scale [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 5931.

PURPOSE:Active surveillance (AS) is an increasingly used strategy for managing localized prostate cancer, yet reliable biomarkers for predicting disease progression remain limited. Low testosterone has been associated with aggressive prostate cancer features at diagnosis, but its role in Grade Group (GG) progression during AS remains unclear. Our objective was to evaluate the association between baseline serum testosterone levels and GG progression in men undergoing AS for localized prostate cancer. MATERIALS AND METHODS:We conducted a retrospective cohort study of 924 men enrolled in an AS program between 2005 and 2024, with a median follow-up of 46.1 months among those who did not progress. Serum testosterone levels were categorized as low if they met a threshold of ≤ 300 ng/dL based on guideline recommendations. Primary outcomes were biopsy progression to GG2 or "extreme" progression to GG3 or higher disease. Multivariable Cox proportional hazards models were used to assess associations between testosterone levels and progression, adjusting for age, PSA density, and biopsy tumor volume. Potential clinically significant confounding by BMI, smoking status, and ethnicity was also considered. RESULTS:Among 924 men, the mean age was 63.6 years (SD 8.1), and the mean PSA density was 0.13 ng/mL2 (SD 0.14). The average baseline testosterone was 394 ng/dL (SD 160.4), with 29.4% (n = 272) of men having a testosterone level ≤ 300 ng/dL. Low testosterone was associated with a statistically significant increased risk of GG3 progression (HR 1.61, 95% CI 1.02-2.54, P = .04). Furthermore, evidence suggested that low testosterone was not associated with risk of GG2 progression (HR 1.25, 95% CI 0.93-1.67, P = .13). Findings were consistent when employing alternative cut points for testosterone and when considering other potential confounders. CONCLUSIONS:These findings suggest that while low testosterone is not clearly associated with moderate progression (GG2), it may increase the risk of higher-grade "extreme" progression to GG3 or higher. Future studies should focus on prospective validation of our findings to elucidate the biological relationship between androgens and prostate cancer progression.

BACKGROUND:Prostate cancer is the most common noncutaneous malignancy among men and disproportionately affects those with low socioeconomic status, particularly men from racial and ethnic minority populations. PURPOSE:This study describes the development of a culturally tailored Mediterranean diet intervention for medically underserved Black and Hispanic men with prostate cancer, using the Intervention Mapping Adaptation (IM ADAPT) framework. METHODS:Conducted at a county safety-net hospital in a large urban area, which serves a population with high medical needs and low socioeconomic status, the project aimed to ensure the intervention was culturally relevant and evidence-based. A collaborative process was used, involving community scientists and patient stakeholders to identify dietary barriers and preferences, while existing interventions were reviewed for cultural fit. Guided by the six steps of the IM ADAPT framework, stakeholder feedback was incorporated throughout the adaptation process. The result was a culturally adapted intervention that included tailored dietary modifications, food provision strategies, and educational materials specifically designed for Black and Hispanic men. RESULTS:Completion of the IM ADAPT steps yielded a culturally adapted Mediterranean diet intervention incorporating tailored dietary recommendations, provision of food, and distribution of educational materials designed to meet the specific needs of Black and Hispanic men. CONCLUSIONS:The IM ADAPT framework enabled the development of a culturally tailored, evidence-based dietary intervention ready for pilot testing among medically underserved men with prostate cancer. This approach may serve as a replicable model for designing interventions for other racial and ethnic minority populations affected by cancer.

XLSX file - 424K, Table 1: Pathways in G x E Analysis, Table 2. Interaction with Obesity at Pathway Level (P< 0.10), Table 3. Interaction with Diabetes at Pathway Level (P < 0.10), Table 4. Interaction with Obesity at Gene Level (P < 0.05), Table 5. Interaction with Diabetes at Gene Level (P < 0.05), Table 6. Interaction with Obesity at SNP Level, Table 7. Interaction with Diabetes at SNP Level, Table 8. Interaction Analysis for Potential Contributing SNPs to Chemokine Signaling Pathway, Table 9. Overlap between Diabetes- and Obesity-related Genes/SNPs.

Mediation analysis is a useful tool to evaluate surrogate endpoints in clinical trials. We propose a novel method, the M-survival learner, for estimating heterogeneous indirect treatment effects in the presence of censored outcomes. The proposed approach enables the identification of interpretable patient subgroups characterized by distinct mediation pathways. To distinguish heterogeneous from homogeneous mediation effects, we introduce a new statistical criterion specifically designed for survival data. The method provides a principled framework for evaluating heterogeneity in surrogate biomarker performance across patient populations, offering evidence to support accelerated approval drug. By explicitly assessing subgroup-specific surrogate validity, the proposed approach addresses key regulatory concerns regarding the reliability of surrogate endpoints. We further establish theoretical properties of the method to justify its statistical guarantees. We apply the approach to data from a Phase III randomized clinical trial of HIV treatment, demonstrating its practical utility in real-world settings. Extensive simulation studies further evaluate and demonstrate its finite-sample performance.

Efforts to translate advances in immunology into anti-cancer immunotherapies have progressed rapidly in recent years. Six antibodies acting on programmed death ligand 1 or programmed death 1 pathways were approved in 75 cancer indications between 2015 and 2021. Several of these therapies were granted accelerated approval for specific cancer indications based on evidence from single-arm phase II clinical trials. In the absence of randomization, however, patient prognosis for progression-free and overall survival may not have been studied under standard chemotherapies for PD-1 and PD-L1 biomarker subpopulations. In 2021, two immunotherapies were withdrawn from accelerated approval applications for treatment of metastatic urothelial carcinoma after randomized phase III trials failed to demonstrate evidence for survival advantage over standard of care. This re-analysis uses digitized data to quantify PD-L1 heterogeneity in chemotherapy response, extending prior meta-analyses by incorporating digitized data and design simulation. The findings of the IMvigor210 (NCT02108652) and IMvigor211 (NCT02302807) trials of atezolizumab are reviewed to elucidate the statistical implications of PD-L1 subpopulation heterogeneity. To place the findings into the context of external evidence, digitization software is used to combine results from journal articles of eleven trials that assigned metastatic urothelial carcinoma patients to the same chemotherapy agents administered in the IMvigor211 control arm. This article defines the extent to which PD-L1 IC2/3 subpopulations appeared to outperform historical expectations in the IMvigor211 study based on external evidence from digitized data. Given the extent of PD-L1 heterogeneity suggested by this analysis, trial simulation is applied to define the probability that IMvigor211 would have resulted in a positive trial based on its actual design and alternative designs that enrolled more IC2/3 patients or had longer durations.

The identification of tumor cells is pivotal for understanding tumor heterogeneity and the tumor microenvironment. Recent advances in spatially resolved transcriptomics (SRT) have revolutionized the way that transcriptomic profiles are characterized and have enabled the simultaneous quantification of transcript locations in intact tissue samples. SRT is a promising alternative method to study gene expression patterns in spatial domains. Nevertheless, the precise detection of tumor regions within intact tissue remains a great challenge. A common strategy for identifying tumor cells is via tumor-specific marker gene expression signatures, which are highly dependent on marker accuracy. Another effective approach is through aneuploid copy number alterations, as most types of cancer exhibit copy number abnormalities. Here, we introduce a novel computational method, called TUSCAN (TUmor Segmentation and Classification ANalysis in spatial transcriptomics), which constructs a spatial copy number variation profile to improve the accuracy of tumor region identification. TUSCAN combines gene information from SRT data and hematoxylin-and-eosin-staining image to annotate tumor sections and other benign tissues. We benchmark the performance of TUSCAN and several existing methods through the application to multiple datasets from different SRT platforms. We demonstrate that TUSCAN can effectively delineate tumor regions, with improved accuracy compared to other approaches. Additionally, the output of TUSCAN provides interpretable clonal evolution inferences that may lead to novel insights into disease development and potential druggable targets.

Surrogate endpoints are widely used in clinical trials to accelerate treatment evaluation, yet their validity may vary substantially across patient subgroups. Although recent advances in heterogeneous causal mediation analysis enable subgroup-specific surrogate evaluation, applying these methods requires substantial expertise in causal inference, statistical programming, and clinical trial methodology, limiting their accessibility to many biomedical researchers. We present SurroPilot, a large language model (LLM)-assisted platform for heterogeneous surrogate endpoint evaluation in clinical trials. Through natural-language interaction, SurroPilot supports the complete analytical workflow, including dataset understanding, data preprocessing, mediator and covariate selection, heterogeneous causal mediation analysis, subgroup interpretation, and automated report generation. To improve the reliability of AI-assisted statistical computing, the platform incorporates a shared context programmerinspector framework for iterative R code correction and automated validation of LLM-generated variable selections. Rather than replacing statistical methodology, SurroPilot integrates LLM with a validated heterogeneous mediation framework, allowing the LLM to assist with analytical reasoning while statistical inference is performed using established causal inference methods. Using the ACTG175 Phase III HIV clinical trial, we demonstrate that SurroPilot provides an end-to-end, reproducible workflow for heterogeneous surrogate endpoint evaluation and substantially lowers the technical barriers to applying advanced causal mediation methods in clinical trial research.

Mediation analysis is a pivotal tool for elucidating the indirect effect of an environmental factor or treatment on disease through potentially high-dimensional omics data, such as gene expression profiles. However, traditional mediation analysis methods tailored for binary outcomes often rely on the rare disease assumption in logistic regression and provide inadequate measures of total mediation effect when multiple mediators have effects in different directions. In this paper, we develop a MEdiation analysis framework in LOgistic regression for high-Dimensional mediators and a binarY outcome (MELODY). It leverages a second-moment-based measure analogous to the R 2 ${R}^{2}$ for linear models to quantify the total mediation effect. We also develop a variable selection procedure for high-dimensional data to reduce bias introduced by non-mediators. Our comprehensive simulations demonstrate the superior performance of MELODY in scenarios with non-rare disease binary outcomes and high-dimensional mediators. We apply MELODY to the Framingham Heart Study of over 5000 individuals to analyze the mediation effects of metabolomics and transcriptomics data on the pathways from sex to incident coronary heart disease.

Supplementary Figure 4. Loss of ADAR1 increases CD8+ T cells infiltration and activation.

Supplementary Figure 1. ADAR1 is upregulated in human pancreatic cancer tissues.

Supplementary Figure 3. Loss of ADAR1 increases CD8+ T cells infiltration.