Objectives Deep learning networks have achieved significant progress in caries diagnosis, but automated localization and numbering of carious teeth in cone-beam computed tomography (CBCT) images remain underexplored. This study aimed to develop a two-stage deep learning framework for tooth detection, numbering, and caries identification in CBCT images, providing a technical basis for automated analysis in support of opportunistic caries screening. Methods This retrospective study included CBCT images from 65 eligible patients. Axial slices were used for model development. For tooth detection, seven classification schemes were designed, and YOLOv3 was compared with Cascade R-CNN. For caries identification, three classification networks (DenseNet169, MobileNet_V2, and ResNet50) were trained and evaluated. Detection performance was evaluated using mean average precision (mAP) and average precision (AP). Identification performance was assessed using balanced accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), Matthews correlation coefficient (MCC), and area under the precision-recall curve (AUC-PR). Results YOLOv3 achieved superior detection performance compared with Cascade R-CNN across all classification schemes (P < .0001). DenseNet169 outperformed the other networks for caries identification. Despite class imbalance, it achieved a balanced accuracy of 0.7414 and an MCC of 0.6074, with high specificity (0.9828) and NPV (0.8976). The integrated two-stage framework showed acceptable overall performance. Conclusion The proposed two-stage framework showed promising performance in detecting, numbering, and identifying caries in CBCT images, supporting the feasibility of adjunctive opportunistic screening on scans acquired for non-caries indications. Clinical Relevance This framework may help prioritize clinician review of CBCT scans for suspected carious lesions, and clinical utility requires external multi-center and prospective validation.

INTRODUCTION:EGFR kinase domain mutations in NSCLC can be classified into four subgroups (classical-like, exon 20 loop insertions, P-loop and αC-helix compressing [PACC], and T790M-like) to effectively predict the response to tyrosine kinase inhibitors (TKIs). These mutations can occur individually or as compound mutations. The frequency, drug sensitivity, and clinical outcomes of compound EGFR mutations, particularly those involving PACC and classical co-mutations, remain incompletely defined. METHODS:We characterized the landscape of single and compound EGFR mutations in 15,851 EGFR-mutant NSCLC samples tested by cell-free DNA. Drug sensitivity was assessed using in vitro Ba/F3 models. Clinical outcomes were retrospectively evaluated from the MD Anderson Cancer Center real-world cohort, the Guardant Health real-world cohort, and a literature review, totaling 1542 patients. RESULTS:Among all EGFR mutations, PACC mutations were found in 9% (1421/15,851) of samples and occurred predominantly as in cis compound mutations (66.2%, 941/1421). Conversely, 84.3% (9576/11,365) of classical mutations and 88.6% (848/957) of exon 20 insertions were mainly single mutations, respectively (p < 0.0001). In vitro testing revealed that compound PACC mutations, including PACC plus PACC and PACC plus classical mutations, exhibited sensitivity profiles similar to those of single PACC mutations, with enhanced sensitivity to second- versus first- or third-generation TKIs. Analysis of retrospective data supported the finding that patients with NSCLC harboring single or compound PACC mutations had improved outcomes with second-generation TKIs compared with first- or third-generation TKIs. CONCLUSION:EGFR PACC mutations primarily occur as in cis compound mutations. Compound PACC mutations exhibit similar patterns of drug sensitivity to those of single PACC mutations.

AIM:This study investigated the transfer of implicit anatomical features from micro-CT to periapical radiographs using fused-rooted mandibular second molars (MSMs) as a model. The objective was to evaluate the feasibility and effectiveness of multimodal transfer learning for the three-dimensional (3D) morphological identification of root canals, and to examine how task complexity influences transfer performance. METHODOLOGY:Fused-rooted MSMs were scanned using high-resolution micro-CT to generate virtual radiographs. Clinically simulated periapical radiographs (CSPRs) were obtained from ex vivo mandibles to reproduce realistic clinical conditions. Based on micro-CT classification, root canals were divided into merging, symmetrical and asymmetrical types. Four convolutional neural network (CNN) architectures (VGG19, ResNet18, ResNet50 and EfficientNet-b5) were trained under three conditions: (1) CSPRs with ImageNet-pretrained CNNs, (2) virtual radiographs with ImageNet-pretrained CNNs, and (3) CSPRs with CNNs pretrained on virtual radiographs. Grad-CAM visualisation was used to interpret model attention, and results were compared with those of four endodontic residents. To reduce task complexity, symmetrical and asymmetrical types were later merged into a "separating" group to generate a two-class classification task. RESULTS:In the three-class task, CNNs pretrained on virtual radiographs achieved an average accuracy of 69.68% (95% CI: 64.61%-74.76%), significantly higher than ImageNet-pretrained models (64.36%, 95% CI: 61.12%-67.61%) and endodontic residents (61.17%, 95% CI: 56.09%-66.25%) (p < 0.05). Grad-CAM visualisation revealed that virtual radiograph-pretrained models concentrated attention on root structures, whereas ImageNet-pretrained networks showed diffuse or misplaced focus. In the two-class task, accuracies were 79.79% (95% CI: 73.30%-86.27%) for CNNs pretrained on virtual radiographs, 73.41% (95% CI: 67.54%-79.27%) for ImageNet-pretrained models and 76.60% (95% CI: 69.28%-83.91%) for residents, with no significant differences (p > 0.05). The overall diagnostic balance improved following transfer learning, indicating better feature representation across classes. CONCLUSIONS:Implicit 3D features extracted from micro-CT-based virtual radiographs can be effectively transferred to CSPRs through transfer learning. This approach enhances CNN interpretability and diagnostic precision in identifying root canal morphology. The benefits of transfer learning are greater for complex, multi-class tasks that require the extraction of intricate morphological features, whereas its effect diminishes in simplified binary classifications. These findings provide a theoretical and experimental foundation for applying multimodal transfer learning to clinical dental imaging.

Abstract Brain metastasis is a major clinical challenge for patients with non-small cell lung cancer (NSCLC), and NSCLC patients with tumors harboring EGFR mutations have an increased risk of central nervous system (CNS) involvement, with up to 50-60% of patients developing CNS metastasis. The process of metastasizing to the brain is a complex multistep process in which tumor cells must cross the blood-brain barrier and adapt to the unique CNS environment. Recent studies in small cell lung cancer (SCLC) suggested that lung cancer cells residing in the brain acquired neural-like transcriptomic programs as compared to primary tumors in the lung. Given that EGFR mutant NSCLC tumor cells can exhibit lineage plasticity through processes including epithelial to mesenchymal transition (EMT) and SCLC or neuroendocrine transformation as part of acquired therapeutic resistance to EGFR inhibitors, we hypothesized that EGFR mutant NSCLC cells may interact with neuronal cells to facilitate CNS metastasis and lineage change. Here, we used in vitro and in vivo models to evaluate the impact of neuronal interaction on EGFR mutant NSCLC cells. We isolated primary murine neurons and co-cultured them with HCC827, HCC4006, and H1975 EGFR mutant NSCLC cells. After two days, immunofluorescent staining for neuronal markers, NeuN and MAP2, showed that EGFR mutant tumor cells co-cultured with neurons upregulated expression of these neuronal markers whereas tumor cells grown alone did not. Our in vivo models of EGFR mutant NSCLC brain tumors supported the enrichment of neuronal, synaptic, and axonal genes in cancer cells exposed to neuronal interaction. Moreover, we performed spatial transcriptomic analysis of human EGFR mutant NSCLC brain tumors paired with primary lung tumors to evaluate alteration of tumor microenvironment (TME) compartments associated with brain metastasis. Together, we propose neural mimicry as a potential mechanism exploited by EGFR-mutant NSCLC cells growing in the brain microenvironment. Our study further sheds light on novel targets for disrupting cancer-neuron interaction to inhibit the growth of brain metastasis in EGFR mutant NSCLC. Citation Format: Sam Song, Tomohiro Takehara, Yan Yang, Monique B. Nilsson, Alissa Poteete, Sherise Desiree Ferguson, Xiuning Le, John V. Heymach. Cancer-neuron interaction initiates neural mimicry and brain colonization in EGFR-mutant NSCLC [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 1076.

Hyperuricemia is a pervasive metabolic disorder defined by dysregulated urate homeostasis, arising from uric acid overproduction, impaired excretion, or both, which elevates serum urate concentrations beyond the physiological saturation threshold. Emerging evidence confirms it as a pivotal risk factor for a spectrum of complications, encompassing gout, renal calculi, and cardiovascular diseases. As its global prevalence escalates, hyperuricemia imposes a considerable burden on public health systems worldwide. Current pharmacological interventions are frequently constrained by adverse effects and inadequate patient tolerance, thus highlighting an urgent need to develop safer and more efficacious regulatory strategies. Natural products have emerged as a promising source for the dietary modulation and adjunctive treatment of hyperuricemia, with their urate-lowering efficacy being increasingly validated. This review provides a comprehensive and critical analysis of recent advances in this field. It systematically summarizes well-characterized anti-hyperuricemic agents derived from terrestrial sources, while placing a particular emphasis on the frontier of marine-derived natural products—an underexplored reservoir with remarkable potential. We elaborate the multi-targeted mechanisms through which these natural products exert their effects, primarily through the inhibition of key enzymes in uric acid biosynthesis and the promotion of renal urate excretion. And also, the body of evidence confirms that these terrestrial and marine natural products offer potent hypouricemic activity coupled with generally favorable safety profiles. Accordingly, these natural products serve as a versatile reservoir of candidate molecules for the development of dietary adjuncts and innovative therapeutic agents tailored to hyperuricemia management.

Colorectal cancer (CRC) remains a leading cause of cancer mortality globally, underscoring the need to identify key molecular drivers. Cell division cycle 20 (CDC20), a regulator of cell cycle progression, is frequently dysregulated in malignancies. Its specific role and mechanisms in CRC pathogenesis are poorly understood, warranting investigation to uncover novel therapeutic targets. In this study, CDC20 mRNA expression was quantified by quantitative real-time polymerase chain reaction (qRT-PCR), while its protein levels were assessed using Western blotting. Cell proliferation was evaluated via the 5-Ethynyl-2’-deoxyuridine (EdU) assay. Cell apoptosis was analyzed by flow cytometry. Migration and invasion capabilities were examined using wound-healing and Transwell invasion assays, respectively. Angiogenesis was assessed using a tube formation assay. To investigate the in vivo effects of CDC20 knockdown, a xenograft mouse model was employed to monitor tumor growth. Flow cytometry was performed to quantify CD206( +) macrophages. Mechanistic studies included chromatin immunoprecipitation (ChIP), dual-luciferase reporter, RNA immunoprecipitation (RIP), and methylated RNA immunoprecipitation (MeRIP) assays to explore GATA binding protein 6 (GATA6) and methyltransferase-like 3 (METTL3) interactions with CDC20. The results showed that CDC20 expression at the mRNA and protein levels was significantly upregulated in CRC tissues in comparison with normal colorectal tissues. Moreover, its protein expression was higher in CRC cells than in human normal colonic epithelial cells. Its knockdown suppressed CRC cell proliferation, migration, invasion, and tube formation, while promoting apoptosis. In addition, CDC20 depletion inhibited tumor growth and immune escape. Transcription factor GATA6 directly activated transcription of the CDC20 gene. Crucially, restoring CDC20 expression counteracted the tumor-suppressive effects of GATA6 knockdown on malignant behaviors and immune escape. METTL3 stabilized CDC20 transcripts via insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2)-mediated mRNA stability. Further, overexpressing CDC20 similarly reversed the inhibitory effects of METTL3 knockdown on CRC cell malignancy and immune escape. Thus, CDC20, upregulated transcriptionally by GATA6 and post-transcriptionally by METTL3, drove CRC progression, angiogenesis, and immune evasion. Targeting CDC20 or its regulators holds significant clinical promise for developing therapies in CRC patients.

Small cell lung cancer (SCLC) is an aggressive neuroendocrine malignancy characterized by rapid onset of chemoresistance and poor clinical outcomes. Following decades of, at best, modest clinical advances, the recent FDA approval of tarlatamab, a DLL3 targeting bispecific T-cell engager (BiTE), alongside unprecedented response rates observed with multiple antibody-drug conjugates (ADCs), have ushered in a paradigm shift towards surface targeting strategies in relapsed SCLC patients. These same agents are demonstrating similar efficacy in more rare high-grade neuroendocrine carcinomas, both pulmonary and extrapulmonary; however, they are being largely tested in unselected populations. While providing much-needed optimism for SCLC patients, resistance, both de novo and acquired, is common and must be better characterized to maximize the potential of these new therapeutic classes. We hypothesize that combinatorial targeting of multiple surface proteins using distinct strategies (i.e., BiTEs, ADCs, etc.) represents a novel way to overcome intratumoral heterogeneity common to relapsed SCLC and enhance antitumor immunity engendered by ADC payloads (i.e., topoisomerase 1 [TOP1] inhibitor). To better define the surfaceome of relapsed SCLC, we performed surfaceome mass spectrometry analysis of SCLC cell lines, naïve and relapsed patient derived xenografts (PDXs), and PDXs treated with frontline chemotherapy until relapse occurred, and identified a number of novel and known surface proteins (i.e., TROP2, HER2, B7H3). Surface targeting strategies against HER2, TROP2, and DLL3 are effective in preclinical models (i.e., cell lines and xenograft models) resistant to other common SCLC therapies (i.e., platinum chemotherapy). In particular, ADCs with TOP1 inhibitor payloads were more effective in models with high SLFN11 and target levels, suggesting that sensitivity requires both surface target expression and SLFN11 positivity for greatest response. Notably, single-cell transcriptional profiling of relapsed patient biopsies revealed mutually exclusive expression of surface genes in distinct cell populations, including senescent, drug tolerant persister cells (DTPCs), representing an unique opportunity to target heterogeneous populations. We show that combination targeting against different surface proteins (e.g., DLL3, TROP2, HER2) using both immune (i.e., chimeric antigen receptor T-cells, BiTEs) and payload-based modalities (i.e., ADCs) was more effective than single-agent targeting in resistant, neuroendocrine-low models. Therefore, intratumoral heterogeneity associated with relapsed SCLC, which limits efficacy of single-agent surface targeting strategies, may be exploited with combinatorial therapies to target resistant cell populations, including DTPCs, using payload and immune-based methods. C. Allison Stewart, Kavya Ramkumar, Runsheng Wang, Yan Yang, Bingnan Zhang, Yuanxin Xi, Lixia Diao, Qi Wang, Alberto Duarte, Ping Li, Azusa Tanimoto, Alejandra G. Serrano, Jody Vykoukal, Mukulika Bose, Loukia G. Karacosta, Luisa Solis Soto, Samir Hanash, Jing Wang, John V. Heymach, Lauren Averett Byers, Carl M. Gay. Combination surface targeting strategies in relapsed small cell lung cancer (SCLC) to overcome intratumoral heterogeneity associated with treatment resistance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2928.

IL-6 modulates expression of NK receptor and their ligands in EGFR mutant tumors.

PURPOSE:Patients with non-small cell lung cancer harboring EGFR mutations typically have significant clinical benefits from EGFR tyrosine kinase inhibitors (TKI) such as osimertinib. However, a residual population of drug-tolerant persister cells (DTPC) inevitably remains, which ultimately gives rise to fully drug-resistant cells (DRC). This study evaluates the activity of EGFR chimeric antigen receptor (CAR)-based therapies in this context. EXPERIMENTAL DESIGN:We developed EGFR CAR T and CAR NK cells and evaluated their antitumor activity against parental cells, DTPC, and DRC in vitro and in vivo. We investigated the mechanisms regulating the sensitivity of DTPC and DRC to CAR T or CAR NK cells, including NK-activating ligands, TGF-β signaling, and EGFR surface levels. Additionally, we developed strategies that included galunisertib treatment and the expression of a dominant-negative TGF-β receptor II in CAR NK cells. RESULTS:DTPC demonstrated increased sensitivity to both EGFR CAR T and CAR NK cells. DRC were relatively resistant to CAR T cells but more sensitive to CAR NK cells. DRC and DTPC had higher levels of natural cytotoxicity triggering receptor-3 and NKG2D ligands, which enhance the effectiveness of CAR NK cells. Elevated TGF-β levels in DRC impaired CAR function, but this was reversed by coexpression of galunisertib or dominant-negative TGF-β receptor II in CAR NK cells. Continued TKI treatment increased EGFR expression on DRC, possibly contributing to the improved killing activity seen with TKI/CAR combinations compared with CAR alone in TKI-resistant cells. CONCLUSIONS:EGFR-directed cellular therapies, particularly EGFR CAR NK cells, demonstrate activity against EGFR-mutant DTPC and DRC in vitro and in vivo, with enhanced activity observed when combined with EGFR TKI or TGF-β pathway blockade.

IL-6 modulates activation state of infiltrating T cells in EGFR mutant tumors.

EGFR-mutant NSCLC tumor cells with acquired resistance to an EGFR-TKI display augmented IL-6 gene expression.

NK cell ligands expression is associated with EMT in EGFR-TKI resistant NSCLC cells

Characterization of EGFR expression

Development of EGFR CAR-NK cells using genetically modified K562 feeder cells

High-grade neuroendocrine carcinomas of the lung (hgNECs), which include small cell lung cancer (SCLC) and pulmonary large-cell neuroendocrine carcinoma (LCNEC), are characterized by poor prognoses, limited treatment options and rapid emergence of resistance. While recent surfaceome-directed approaches targeting DLL3, SEZ6, and TROP2 using chimeric antigen receptor T cells (CAR-Ts), bispecific T cell engagers (BiTEs) and antibody drug conjugates (ADCs) show considerable promise in hgNECs, novel strategies to target the heterogenous resistant disease are needed. AXL, a TAM family receptor tyrosine kinase, is known to mediate resistance to chemotherapy, radiation and targeted therapies in SCLC and other cancers, through its roles in epithelial to mesenchymal transition (EMT), DNA damage repair and replication stress tolerance. We hypothesize that targeting cell surface AXL would provide an effective therapeutic strategy in hgNECs, especially for relapsed disease. We analyzed the transcriptomic (bulk and single-cell RNA sequencing) and proteomic (immunohistochemistry) expression profiles of AXL in treatment-naïve and relapsed SCLC, LCNEC and other extrapulmonary hgNECs patient tumors. High levels of AXL expression were seen in distinct tumor subsets. Notably, AXL-high tumors exhibited molecular signatures associated with resistance (higher EMT scores, lower replication stress and cisplatin response). To evaluate AXL as a viable surface target in hgNECs, we generated a preclinical CAR T cell against AXL. Anti-AXL CAR T cells were effective at inducing cytotoxicity in AXL-expressing SCLC and LCNEC cell lines. Similarly, an anti-AXL ADC also showed selective and potent cytotoxicity in AXL-positive models. These findings establish AXL as both a valuable biomarker and a promising target for the drug-resistant hgNECs, supporting the development of AXL-directed therapeutics such as anti-AXL CAR T cells and anti-AXL ADCs. Kavya Ramkumar, C. Allison Stewart, Yan Yang, Bingnan Zhang, Qi Wang, Yuanxin Xi, Runsheng Wang, Alejandra G. Serrano, Luisa M. Solis Soto, Jing Wang, John V. Heymach, Carl M. Gay, Lauren A. Byers. AXL-directed surface targeting approaches in recalcitrant pulmonary high-grade neuroendocrine carcinomas [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2931.

Fueled by rapid advances in gene editing, synthetic biology, artificial intelligence, regenerative medicine, and brain-computer interfaces, biotechnology is approaching a transformative era often referred to as biotechnological singularity. CRISPR-based gene editing has revolutionized genetic engineering, enabling precise modifications for treating hereditary diseases and cancer. Synthetic biology facilitates sustainable biomaterial production and innovative therapeutic applications. Artificial intelligence accelerates drug discovery, enhances diagnostic accuracy, and personalizes treatment through deep learning models. Driven by stem cell research, regenerative medicine offers promising avenues for reversing aging and treating degenerative diseases. Brain-computer interfaces merge human cognition with technology, enabling direct neural control of prosthetics and expanding human-machine interactions. These breakthroughs, however, raise ethical, regulatory, and societal concerns, including equitable access, biosecurity risks, and the implications of human enhancement. The convergence of biological and computational technologies challenges traditional boundaries, necessitating comprehensive governance frameworks. By embracing responsible innovation, society can harness these advancements for transformative health interventions, environmental sustainability, and extended longevity. The realization of biotechnological singularity depends on interdisciplinary collaboration among scientists, policymakers, and the public to ensure that progress aligns with the well-being of humanity and ethical considerations.

OBJECTIVE:Cone-beam computed tomography (CBCT) is used extensively in dental practice but has limited spatial resolution for visualising fine root canal structures. Micro-computed tomography (micro-CT) offers superior resolution but is unsuitable for clinical use. This study investigated the possibility of enhancing CBCT resolution through deep learning-based super-resolution, using paired micro-CT images as the ground truth. METHODS:Two architectures, Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) and Hybrid Attention Transformer (HAT), were trained and evaluated. An edge loss function combining Gaussian and median filtering with Sobel edge detection was introduced to improve structural detail. CBCT and micro-CT images from 48 extracted human teeth were processed into matched datasets. Performance was evaluated using peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), expert visual appraisal, and three-dimensional reconstructions. RESULTS:Both ESRGAN_edge and HAT_edge significantly outperformed bicubic interpolation and their non-edge-loss counterparts. Subjective ratings indicated that ESRGAN_edge and HAT_edge approached micro-CT quality. Three-dimensional reconstructions confirmed improved anatomical accuracy of pulp chamber and root canal structures, with ESRGAN_edge showing the greatest overlap with micro-CT. Clinical CBCT testing demonstrated that the trained models enhanced root canal clarity, although artefacts in crown regions require further refinement. CONCLUSIONS:Micro-CT-guided super-resolution, particularly with edge optimisation, substantially improved CBCT diagnostic utility in endodontics CLINICAL SIGNIFICANCE: The super-resolution models investigated in the present work achieve acceptable results in enhancing the resolution of the roots of teeth in clinical CBCT scans. Edge-aware super-resolution deep learning models hold promise for clinical dental imaging.

EGFR downregulation in EGFR-TKI resistant NSCLC cells and associated induction of exhaustion markers in CAR-T cells

Development of IL-6 knockout EGFR-mutant GEMM tumors.