JCO Clinical Cancer Informatics

PURPOSE: To develop and validate an artificial intelligence-enabled platform that converts unstructured cancer trial eligibility criteria into structured queries and quantifies trial eligibility across advanced/metastatic cancer trials. METHODS: We downloaded actively recruiting US interventional treatment trials for advanced/metastatic breast cancer, colon cancer, and non-small cell lung cancer from ClinicalTrials.gov. Medical oncologists created 24 synthetic patient vignettes. A large language model converted trial eligibility criteria into Structured Query Language (SQL) code and patient information into structured records, enabling automated matching. Cancer details and treatment history were considered, but not laboratory results or comorbidities. Validation included physician editing of generated eligibility code for 30 trials, and blinded physician eligibility assessment for five trials. We then evaluated how age, ECOG performance status, sex, and ZIP code affected the number of eligible trials. RESULTS: Of 833 candidate trials, 746 met inclusion criteria. In physician review of 30 trials, edits to generated SQL did not change any of 720 trial-patient eligibility determinations for 24 synthetic patients. In blinded validation across 120 trial-patient pairs, automated matching achieved 97% accuracy. Across synthetic patients, eligible trials ranged from 31 to 258 when there were no geographic restrictions. Eligibility decreased markedly with worse performance status and with geographic restriction (both p<0.001). Later-phase, randomized, and molecularly selective trials had fewer eligible patients. CONCLUSION: AI-based structuring of trial eligibility criteria can support accurate, scalable measurement of potential cancer trial eligibility. In this demonstration, performance status, geography, and age were major determinants of eligibility across the active metastatic trial landscape.

Background. The European Society for Medical Oncology (ESMO) Scale for Clinical Actionability of molecular Targets (ESCAT) ranks genomic alterations by the evidence supporting the predictive value of the molecular target for response to targeted therapies. No openly available, systematically curated set of standard care biomarkers mapped to the ESCAT framework exists to support clinical decision-making or harmonize biomarker interpretation. Methods. We mapped all OncoKBTM Level 1 biomarkers to ESCAT tiers using evidence cited by OncoKBTM, excluding abstract-only data. Eight board-certified oncologists and hematologists independently assigned ESCAT tiers, with discrepancies resolved through structured consensus meetings. Recurring evidence scenarios that did not correspond to any existing ESCAT tier informed a set of a priori defined modifications, which were subsequently applied to biomarkers that could not be classified using native ESCAT criteria. Results. Of 188 OncoKBTM Level 1 biomarkers, 16 were excluded due to abstract-only evidence. Using native ESCAT criteria, 51% of the remaining biomarkers were classified as Tier 1, 3% Tier 2, 18% Tier 3, 6% Tier X and 22% could not be assigned to any tier. Applying the modified ESCAT criteria resolved all previously unclassifiable biomarkers and increased Tier 1 assignments to 73%. Inter-rater reliability (Krippendorffs alpha) was moderate (0.586) and 62% of classifications required consensus discussions. Comparison with ESCAT tiers reported in ESMO Clinical Practice Guidelines showed improved concordance when using the modified criteria. Conclusions. The native ESCAT criteria are highly stringent, resulting in many FDA-recognized, clinically validated biomarkers that are currently assigned level 1 by OncoKBTM not mapping to any existing tier. Our predefined modifications improved alignment with OncoKBTM Level 1 designations and with published ESMO clinical practice guidelines. The mapped set of standard care biomarkers are provided on the OncoKBTM website, offering a practical resource that harmonizes ESCAT tiers of evidence with a widely adopted levels of evidence schema.

Background: Structuring oncology clinical notes into registry-grade variables is essential for research and care but remains labour-intensive and error-prone. Objective: To develop and evaluate a privacy-preserving large language model pipeline for oncology registry abstraction in a real-world clinical setting. Methods: We deployed an open-source Meta Llama 3.3 70B-based pipeline to extract over 50 variables from 6,700 oncology notes at a cancer centre in Singapore. Data were de-identified locally using a Hide-In-Plain-Sight approach, ensuring no identifiable data left hospital infrastructure. Performance was assessed on 200 randomly sampled notes with adjudicated ground truth. A structure-aware framework classified outputs as correct, missing, spurious, or incorrect. Results: F1 scores were high across variables, including diagnosis (97.2%), histology (95.8%), stage (92.6%), biomarkers (91.4%), and treatments (88.1%). Transferability testing on 50 external notes showed strong performance for core variables. Conclusions: Privacy-preserving LLMs can achieve near-human-level accuracy for oncology abstraction, with structure-aware evaluation enabling more clinically meaningful assessment. Keywords: Oncology Registry Abstraction, Privacy-Preserving Deployment, Clinical Information Extraction, Structure-Aware Evaluation, Large Language Models, Template-Filling Metrics

Background: Circulating tumour DNA (ctDNA) liquid biopsy is now established across oncology for early cancer detection, minimal residual disease surveillance, and treatment monitoring. Detection thresholds for all current ctDNA assays are derived empirically through receiver operating characteristic analysis on training cohorts - a statistically valid but theoretically uninformed approach that does not specify the minimum detectable tumour fraction given assay technical characteristics, nor identify when increasing sequencing depth ceases to provide additional clinical information. Methods: We model ctDNA detection as a binary hypothesis testing problem with Binomial-distributed mutant allele counts against a sequencing error noise floor. The Neyman-Pearson lemma is applied to derive the uniformly most powerful detector and the minimum detectable tumour fraction in closed form. The sequencing assay is modelled as a binary symmetric channel and Shannon channel capacity is calculated. Empirical validation uses n=61 data points extracted from five published peer-reviewed analytical validation studies across five independent institutions in the US and EU (2018 - 2025): Yu et al. 2022, Stetson et al. 2018, Frydendahl et al. 2023, Northcott et al. 2024, and Cheng et al. 2025. Results: The minimum detectable tumour fraction is derived in closed form as f_min approximately equal to (z_alpha + z_beta) multiplied by the square root of (epsilon divided by N), where N is sequencing depth, epsilon is the platform error rate, and z_alpha, z_beta are standard normal quantiles at the specified false positive and false negative rates. Shannon channel capacity is C = 1 minus H(epsilon) bits per read, where H(epsilon) is binary entropy. Empirical validation yields 84.3% agreement for single-locus assays. Discordance for multi-locus tumour-informed assays (NeXT Personal, duplex WGS) is consistent with the single-locus model scope and identifies the principal theoretical extension required. Conclusions: This framework provides the first formal Neyman-Pearson optimality proof for ctDNA detection, a closed-form detection limit, and a platform-independent efficiency metric for NHS and regulatory standardisation. Keywords: circulating tumour DNA; liquid biopsy; Neyman-Pearson detection; Shannon channel capacity; sequencing depth; limit of detection; minimal residual disease; signal detection theory

Background: Preparing tumor board patient summaries is time intensive. Large-language-model based systems may automate summarization but require real-world evaluation prior to clinical use. We performed an exploratory retrospective evaluation of the Microsoft Healthcare Agent Orchestrator (HAO), deployed in a Mayo Clinic controlled staged environment, to generate tumor board-style patient summaries from retrospective Electronic Health Record (EHR) notes. Methods: HAO generated summaries for breast, hepatobiliary, and neuro-oncology tumor board cases using up to the most recent 1,000 clinical notes. Clinician reviewers evaluated outputs via REDCap surveys across perceived factuality, completeness, clarity/conciseness, temporal cohesion, comparative performance, safety, and clinical utility (0-4 Likert scale). Reviewers were permitted to query the HAO chat interface to address missing details. Automated factuality was assessed using TBFact (bidirectional entailment), reporting precision and recall against available reference summaries. Results: Among 57 survey responses from 5 different physicians, mean scores exceeded 2.8 across domains, with medians of 3 for most axes. In an exploratory comparison, oncology fellows required less time to review HAO-generated summaries than to manually generate patient summaries (mean difference 13.57 minutes per patient, p<0.001), although this difference may be influenced by prior familiarity with the same cases; 96% of survey responses indicated that HAO would save time. TBFact evaluations showed higher recall than precision across domains, consistent with broad capture of reference content alongside additional content that was not present in gold-standard summaries. Attribution was viewed favorably but showed issues with primary-source specificity and link reliability. Conclusions: In a controlled Mayo environment, HAO demonstrated moderate performance and was associated with reduced review time for tumor board preparation. These findings are promising but preliminary and do not establish clinical safety, noninferiority to manual review, or readiness for routine clinical use. Limitations, including verbosity, specialty-specific content gaps, and inconsistent attribution, highlight the need for iterative refinement and further evaluation.

PurposeMachine learning (ML) models for omics-based cancer prognosis are often trained on data from predominantly European-ancestry populations, producing biased predictions for other populations and undermining equitable genomic medicine. Existing fairness benchmarks mainly focus on outcome parity rather than predictive performance parity across populations. Public benchmark resources are needed for systematically detecting and mitigating such performance disparities in multi-population cancer prognosis. MethodsWe developed Equitable Health Intelligence (EHI, https://ehiportal.org), an open-source benchmark of multi-population ML for omics-based cancer prognosis. EHI contains 1,475 ML tasks across 40 cancer/pan-cancer types, 4 omics feature sets, 4 clinical endpoints, 5 event-time thresholds, and 3 data-disadvantaged population (DDP) groups relative to a majority European Ancestry population group. Deep neural network models are trained under three multi-population ML schemes (Mixture, Independent, and Transfer Learning), with Naive Transfer included as a no-adaptation control, comprising a total of 10,325 ML experiments. ResultsThe EHI platform provides an interactive environment with visualization and exploratory tools for users to inspect predictive performance disparities between the majority European-ancestry group and data-disadvantaged populations, evaluate the extent to which transfer learning mitigates these disparities, and examine the impact of feature engineering methods across cancer types, omics features, and clinical endpoints. ConclusionEHI is an open, interactive, and extensible benchmark for identifying and addressing performance disparities in multi-population ML for omics-based cancer prognosis. It provides a foundation for a growing ecosystem of methods targeting ML performance disparities arising from biomedical data inequality and population-level distribution shifts, thereby advancing equitable AI in precision oncology.

Clinical chart abstraction extracts structured patient variables from longitudinal clinical notes but is labor-intensive and difficult to scale. We evaluated LLM agents for question-guided chart review using lung cancer molecular testing guideline concordance as a use case. Two configurations were compared: (1) sequential note review using metadata and chronology, and (2) the same framework augmented with keyword-based note search. Gold-standard labels were established by human annotators. The search-enabled agent achieved higher accuracy (92.4% vs. 83.5%) and reduced errors by more than half (41 vs. 89) by retrieving evidence from long, heterogeneous note histories. In guideline concordance evaluation, most determinate patient-rule assessments were concordant (80.7%), while most apparent non-concordance reflected missing molecular testing documentation rather than documented care deviations. These results suggest tool-augmented LLM agents can approximate key aspects of human chart review and support scalable information extraction from longitudinal clinical documentation.

Background: Molecular medicine has made genetic diagnostics crucial for rare diseases, but the majority of patients remains without diagnosis even after state-of-the-art assessment. Standardized systems for integrating clinical features, such as the Human Phenotype Ontology (HPO), offer assistance, but are often insufficiently detailed and fail to capture crucial clinical parameters such as age at onset, longitudinal changes in symptoms, detailed characteristics of a clinical symptom, or the absence of a feature. Results: We present Genosolver an integrated workflow that utilizes machine learning to address this bottleneck. Using Large Language Models (LLMs) and Large Reasoning Models (LRMs) on unstructured clinical notes and electronic health care data, we generate a workflow that unifies phenotype extraction, generates differential diagnosis, and prioritizes genetic variants from genome data. We evaluated the performance on 233 previously genetically solved cases, where Genosolver ranked the causative gene first in 72% of cases and in 94% of cases in the top 10 gene list, outperforming the existing benchmarking tool Exomiser by 9%. Semi-automated reanalysis of 1,875 unsolved rare disease cases yielded an additional diagnostic rate of 1.7%. Incorporating rich, unstandardized clinical narratives substantially enhanced model performance beyond HPO-only inputs and demonstrated competitive results using data security compliant local models. Conclusion: Integrating unstandardized clinical data with local LLMs and reasoning offers a scalable, data-secure workflow that increases molecular diagnoses in rare diseases.

Our study evaluated whether a deep learning auto segmentation model combined with machine learning triage can streamline radiotherapy clinical trial quality assurance (QA). We analyzed 107 stereotactic ablative radiotherapy (SABR) cases from a multi-institutional phase II clinical trial of neurovascular sparing prostate SABR, focusing on physician contours of the internal pudendal artery (IPA) as a novel organ-at-risk with substantial interobserver variability. Contours were scored by the trial principal investigator as Per-Protocol or Minor Deviation/Unacceptable. We applied a deep learning model for IPA auto-segmentation. Agreement between human and AI contours was then quantified using 14 overlap, distance, and surface metrics, and a supervised classifier was trained on these metrics to flag clinical trial protocol deviations. While AI segmentation achieved only modest geometric accuracy with mean Dice similarity coefficient of 0.446 and 95th percentile Hausdorff distance of 14.23, when incorporating all 14 metrics, a machine learning classifier yielded AUROC of 0.836, flagging all Minor Deviation/Unacceptable cases with 100% sensitivity on the 27 case hold-out set with 6 false positives and no false negatives. AI segmentation combined with metrics-based machine learning can triage protocol deviations within a multi-institution radiotherapy clinical trial, supporting prospective evaluation of AI-assisted trial QA.

Introduction In analyses of time-to-event data, clinical characteristics can have non-linear impacts on survival outcomes, and understanding this dynamic behavior is crucial for producing real-world evidence (RWE). Nonetheless, estimating these dynamic effects is inherently challenging when utilizing real-world data (RWD), especially since sharing individual-level patient data (IPD) is heavily restricted due to regulatory limitations. Additionally, computational difficulties are exacerbated by the high dimensionality, inter-dependency, rarity, sparsity, and scarcity of features. While data augmentation through collaboration across multiple sites might address these challenges, such collaboration is often infeasible and hindered by regulatory measures that protect patient privacy, thereby preventing the sharing of IPD between sites. Objectives To address this challenge, we propose a privacy-preserving regularized algorithm that eliminates the necessity of aggregating any protected health information across sites. This algorithm employs a penalized federated additive model utilizing piecewise exponential survival (FAMES) data and estimates non-linear effects of features while accounting for non-varying confounding effects. The model is flexible and can accommodate both multiple and multivariate smooth effects simultaneously. Methods The proposed model transforms survival data into a piecewise exponential data (PED) structure and casts the semi-parametric optimization problem into a generalized additive modeling framework assuming Poisson distribution. The model uses orthonormal splines to approximate non-linear effects and incorporates L2-norm based penalty terms to control the smoothness and goodness-of-fit of these effects. The algorithm is optimized using site-specific aggregated summary statistics and is solved iteratively through the Newton-Raphson method. Results The model is employed to assess the smooth effects of clinical features, such as age and numeric laboratory values, on overall survival using RWD from approximately 874 newly diagnosed Acute Myeloid Leukemia (AML) patients treated at seven distinct sites in the United States. The model exhibited non-linear smooth effects for lactate dehydrogenase, platelets, and others underscoring their strong association with disease prognosis. The model demonstrates a lossless property, providing estimates of smooth and fixed effects that are comparable to those derived from the pooled PED. Additionally, the inference of parameters for testing the nullity of effects remains consistent. This model is communication-efficient, necessitating roughly twelve rounds of communication across sites. Conclusion We anticipate that this model can facilitate multisite collaboration and enable smaller sites to participate in generating and validating RWE, especially for rare diseases. While the model was applied within the context of AML, it is disease-agnostic and can be implemented in any other clinical context and across various sites globally without losing any generality.

Clinical AI systems have achieved strong predictive performance; however, prediction accuracy is not sufficient for clinical safety. Retrieval-augmented generation (RAG) improves factual accuracy, and general-purpose LLM guardrails constrain surface-level output safety, but these mechanisms do not govern the inferential gap between available clinical evidence and permissible clinical claims. We propose Evidence-Graded Decision Authorization (EGDA), a framework that separates evidence extraction, sufficiency assessment, and claim-level authorization through domain-specific rules. In a controlled experiment using 60 breast cancer decision-snapshot cases (1,260 system outputs across three arms evaluated by LLM-as-Judge with expert calibration), EGDA reduced the unjustified inference rate to 8.0% (vs. 48.7% for unconstrained LLM and 47.7% for RAG; risk difference vs. unconstrained -40.7%, 95% CI -46.9 to -34.0, p < 0.001), raised the appropriate refusal rate to 95.0% (vs. 56.9% and 56.9%; risk difference vs. unconstrained +38.1%, 95% CI +31.3 to +44.5, p < 0.001), and achieved the highest factual correctness at 96.4% (vs. 69.8% and 74.5%). An unexpected finding was that retrieval-augmented generation without an authorization gate failed to reduce unjustified inference relative to the unconstrained baseline (47.7% vs. 48.7%, p = 0.870) and produced no improvement in appropriate refusal (56.9% vs. 56.9%, p = 1.0), showing that information supply alone is not sufficient for inferential governance. We argue that domain-specific, evidence-graded reasoning governance should serve as a deployment reference standard for safety-critical clinical AI.

Background: Accurate extraction of Human Phenotype Ontology (HPO) terms from clinical notes is essential for variant prioritization and genetic diagnosis. Large language models (LLMs) often struggle to balance precision, hallucination avoidance, and ontology mapping accuracy, and prior work has shown that retrieval-based grounding can improve performance for individual models. We hypothesized that real-time ontology grounding through external tools would improve these metrics across heterogeneous LLMs, and we evaluated the Model Context Protocol (MCP), a standardized open framework for integrating external tools, as a vendor-agnostic mechanism for delivering such grounding. Methods: Five LLMs (Claude Sonnet 4.5, GPT-5.1, Gemini 2.5 Pro, Grok 4.1, and Qwen3 30B) extracted HPO terms from four synthetic clinical genetics notes under two conditions: baseline ("No Tools," internal knowledge only) and tool-augmented ("With Tools"), with real-time HPO retrieval delivered through MCP for models with native support and through functionally equivalent native tool-calling interfaces otherwise. Each model performed [&ge;]50 runs per note per condition (>2,000 total runs). Performance was evaluated using Precision, Recall, and F1-score. Outputs were manually adjudicated to classify mapping errors and hallucinations. Results were benchmarked against a commercial EHR-based HPO extraction tool. Results: Tool augmentation significantly improved performance across all models. Mean aggregate F1-score increased from 0.46 (SD 0.22) in the baseline condition to 0.72 (SD 0.15) with tools (p < 0.001). Mapping Error Rate decreased from 40.9% to 7.8% (p < 0.001), and Precision increased from 56% to 90%. Performance gains were observed across all model families, including the open-weight Qwen3 model (F1 0.11[-&gt;]0.50). For inferred phenotypes, F1 improved from 0.20 to 0.34 (p < 0.001) without a significant increase in hallucination rate (p = 0.08). Compared with the commercial benchmark, tool-augmented LLMs achieved higher F1-scores and substantially greater recall for inferred phenotypes. Conclusions: Real-time ontology grounding substantially improves HPO extraction across diverse LLMs by reducing mapping errors and enhancing phenotype inference. The Model Context Protocol provides a standardized, interoperable mechanism for delivering such grounding, supporting reproducible, vendor-agnostic deployment of clinical LLM pipelines in genomic medicine.

Background: Collaborative research programs increasingly require infrastructure capable of integrating heterogeneous participant, sample, and experimental data while meeting evolving research needs. Existing tools, including clinical EHRs, REDCap, generic research information management systems, and bespoke database builds, were not designed to operationalize project-specific data models. The Institute for Clinical and Translational Science (ICTS) at the University of California, Irvine (UCI) ICTS-Dashboard fills this need by providing a general purpose research information management system. Methods: We describe the ICTS-Dashboard, built as an open-source, schema-driven platform in which database structure, server-side validation, representational state transfer application programming interfaces (REST APIs), web-based forms, and reproducible exports are all generated from a single versioned java script object notation (JSON) Schema set. The backend is implemented in Django, Django REST Framework, and PostgreSQL; the frontend in React. We instantiate the platform with the Genomics Research to Elucidate the Genetics of Rare Diseases (GREGoR) Data Model and extend it with two case studies: a locally developed biobank table for biospecimen logistics, and an embedded adaptation of the RAG-HPO retrieval-augmented phenotype curation tool. Results: The ICTS-Dashboard deployed at the UCI-GREGoR site supports 37 schema-derived tables and 250 documented API endpoints. It holds metadata for 2,558 participants, 1,237 families, 5,517 biobank entries, 2,466 sequenced biospecimens, and 289 genetic findings, and supports quarterly external data submissions regenerated directly from the database. The biobank extension adds entities the consortium does not standardize while preserving foreign-key linkage to rare disease records; the RAG-HPO module adds curator-mediated phenotype normalization against 19,389 indexed HPO terms. Both were integrated without modifying the GREGoR data model. Conclusion: A version-controlled, machine-readable data model can serve not only as a data sharing standard but as the operational backbone of a research program when paired with schema-governed tooling. The Dashboard's architecture is not intrinsic to a data model or to rare disease; any collaborative research program with a structured, versioned model can adopt the same pattern to reduce implementation overhead and improve reproducibility, harmonization, and findable, accessible, interoperable, and reproducible (FAIR)-aligned accessibility.

Background: The development of personalised mRNA cancer vaccines holds considerable promise for oncology, yet a significant translational gap persists between neoantigen identification and the selection of therapeutically impactful targets. Current approaches predominantly prioritise human leukocyte antigen (HLA) binding affinity and immunogenicity, often overlooking the systems-level biological context of the target. This can inadvertently favour immunogenic but biologically peripheral peptides that exert limited influence on tumour signalling networks, thereby constraining vaccine efficacy. Furthermore, mRNA therapeutics must satisfy additional design requirements, including favourable codon usage and favourable secondary-structure stability, which directly affect in vivo translation and half-life. A unified computational framework that integrates neoantigen discovery with network biology is therefore critically needed. Results: Here, we present PimRNA, a Priority index (Pi)-centric computational medicine framework that bridges this gap by unifying neoantigen identification, mRNA sequence optimisation, and gene interaction network analysis. First, high-confidence tumour-specific HLA class I and II neoantigenic peptides are identified from paired tumour-normal genomic and tumour transcriptomic data using NeoDisc. Second, the coding sequences of these peptides are optimised for stability and translational efficiency with LinearDesign, yielding a core set of neoantigen-encoding mRNAs. Third, a random walk with restart algorithm is applied to a knowledgebase of gene interactions to identify peripheral genes exhibiting significant network connectivity to core genes, generating a gene-predictor matrix in which each gene is assigned an affinity score reflecting its network proximity to immunogenic neoantigens. These scores are consolidated into a single, unified priority rating (0-5) for each gene, followed by subnetwork analysis that reveals therapeutically relevant gene modules. Application of PimRNA to breast cancer and melanoma datasets demonstrates that it successfully selects high-confidence immunogenic neoantigen candidates embedded within biologically meaningful tumour-specific networks. Conclusion: PimRNA provides a systems biology foundation for mRNA vaccine design, moving beyond isolated immunogenicity to prioritise targets that are both highly presented and central to tumour-relevant biological networks. This framework offers a generalisable strategy for the rational discovery and prioritisation of mRNA therapeutics, significantly advancing the field of computational medicine towards personalised cancer vaccines.

Introduction Coding tumor diagnoses from free-text clinical documentation currently requires substantial manual effort. Promising approaches for automating this process include large language mod-els (LLMs), embedding models, and retrieval-augmented generation (RAG). While previous studies often focus on a single method, we directly compare these approaches on a real-world dataset of tumor diagnosis descriptions to assess their strengths and limitations. Methods We evaluated nine different embedding models using similarity search and embedding-based classification, as well as LLM-based coding, with and without RAG, on a real-world dataset of 2,024 unique German tumor diagnosis descriptions labeled with ICD-10 and ICD-O topography codes. The retrieval knowledge base was constructed exclusively from stand-ardized Alpha-ID, ICD-10-GM, and ICD-O-3 classifications. Performance was assessed for exact (full-code) and partial (three-character) code prediction. For RAG, we evaluated base and fine-tuned versions of Llama 3.1 8B and Llama 3.3 70B. Results Qwen3-Embedding-8B, the largest embedding model, yielded the best results. It achieved 47.8% exact-match and 72.1% partial-match accuracy for ICD-10 coding with classification, and 42.7% exact-match and 73.5% partial-match accuracy for ICD-O coding with similarity search. The other embedding models, including medically specialized ones, showed varied but lower performance. RAG improved base LLM perfor-mance and outperformed embedding-based approaches on partial-match accura-cy (80.6% partial-match accuracy for ICD-10 and 75.0% for ICD-O with Llama 3.3 70B), but not on exact-match accuracy. Conclusion A direct comparison with embedding-based approaches is essential to determine whether the additional effort of RAG is justified. The strong variation in performance also highlights the importance of model selection. Further advances in embedding-based methods, potential-ly supported by larger and more diverse training data, may offer a promising direction for future work.

Importance Prognostic tools beyond staging are needed to guide treatment and counseling in head and neck squamous cell carcinoma (HNSCC). Objective To develop and externally validate a machine learning model predicting survival in advanced HNSCC using routinely collected clinical and biomarker data. Design, Setting, and Participants Retrospective, multi-institutional cohort study including 2,385 patients with stage III-IV HNSCC diagnosed from 2012-2022 in the University of California Health Data Warehouse (UCHDW). Patients were randomly split into training (n = 1,908) and test (n = 477) sets. Partial external validation used 7,749 patients from the Surveillance, Epidemiology, and End Results (SEER) registry (2010-2020). Exposures Demographic, tumor, treatment, comorbidity, and biomarker variables recorded at or before diagnosis. Main Outcomes and Measures The primary outcome was all-cause mortality within 70 months. Cox proportional hazards models included all predictors. Discrimination was assessed with Harrell's concordance index (C-index), calibration with predicted vs observed survival, and stratification with Kaplan-Meier curves. A Random Survival Forest (RSF) was trained for benchmarking and interpretability using Shapley Additive exPlanations (SHAP). Results Among 2,385 patients in UCHDW (median age, 63 years; 29.0% mortality), the Cox model achieved a C-index of 0.735 in the internal test set. Risk quartiles showed clear separation on Kaplan-Meier curves (log-rank p < 0.0001). In the SEER cohort (n = 7,749), where only demographic, staging, subsite, and treatment variables were available, the reduced Cox model achieved a C-index of 0.688, with calibration showing modest underestimation of survival in high-risk groups. Age, T stage, Charlson Comorbidity Index, neutrophil-to-lymphocyte ratio, and platelet count were among the strongest predictors, while surgery was associated with improved survival. The RSF achieved a C-index of 0.758 internally, with SHAP highlighting nonlinear effects of albumin, BMI, and inflammatory markers. Conclusions and Relevance A machine learning model using routine clinical and biomarker data demonstrated good prognostic performance in advanced HNSCC, with partial external validation. Such approaches may support individualized survival estimates, risk stratification, and treatment discussions, but broader validation is required before clinical adoption.

BackgroundAcute myeloid leukemia (AML) remains the most lethal acute leukemia in adults, with 5-year overall survival below 32% despite recent advances including venetoclax-, FLT3-, IDH1/2-, and Menin-targeted therapies. Clinical outcomes remain highly heterogeneous across patients, highlighting the need for robust molecular biomarkers capable of improving prognostic precision. MicroRNAs (miRNAs) are critical regulators of hematopoietic differentiation, apoptosis, and therapeutic resistance and are differentially expressed across AML subtypes. However, their clinical translation has been limited by high dimensionality, feature redundancy, and relatively small cohort sizes. MethodsWe developed and evaluated the AML Survival Estimator (AMLS), an inheritable bi-objective combinatorial genetic algorithm integrated with support vector regression (SVR), using TCGA-LAML miRNA expression profiles (n = 156). AMLS was benchmarked against ten widely used machine-learning approaches, including penalized regression, tree-based ensembles, support-vector regression, k-nearest neighbors, and multilayer perceptron models. Performance was assessed using stratified cross-validation with Pearson correlation (R), Harrells concordance index (C-index), and mean absolute error (MAE). Functional characterization of the derived miRNA signature was performed through consensus target integration followed by pathway enrichment, gene ontology analysis, network reconstruction, and Kaplan-Meier risk stratification. ResultsAMLS achieved superior prognostic performance with pooled out-of-fold metrics of Pearson R = 0.86, C-index = 0.788, and MAE = 7.49 months, substantially outperforming all comparator models. Restricting analyses to the AMLS-derived 28-miRNA signature improved all baseline learners by approximately 2-4-fold, with the multilayer perceptron achieving R = 0.674; however, none matched the native AMLS framework, indicating that the evolutionary optimization strategy contributes predictive information beyond feature selection alone. The prognostic signature included biologically established AML-associated miRNAs, including hsa-miR-191, hsa-miR-29c, hsa-miR-125b, hsa-miR-148a, hsa-miR-15b, hsa-miR-10b, and hsa-miR-30c, linked to DNA methylation, apoptosis, cell-cycle regulation, and oncogenic Wnt/MAPK signaling pathways. Functional analyses demonstrated significant enrichment of canonical AML-associated pathways, including p53, PI3K-AKT, TGF-{beta}, JAK-STAT, FoxO, and hematopoietic lineage signaling. ConclusionsOur findings demonstrate that evolutionary learning integrated with SVR can recover a compact and biologically interpretable miRNA prognostic signature that substantially outperforms conventional machine-learning approaches for AML survival prediction. The identified miRNA network converged on key leukemogenic pathways involved in apoptosis, cell-cycle regulation, and oncogenic signaling, supporting both the biological relevance and prognostic utility of the framework. Given the minimally invasive and quantitatively scalable nature of miRNA profiling, this approach may provide a practical molecular adjunct for improving prognostic assessment and precision medicine strategies in AML. Abstract FigureSchematic overview of the AMLS framework. Left: acute myeloid leukemia, a clonal hematological malignancy with persistent prognostic heterogeneity. Middle: AMLS couples an evolutionary learning-based feature selection algorithms to support vector regression for miRNA-based survival modeling. Right: AMLS recovers a 28-miRNA prognostic signature that predicts overall survival with Pearson R = 0.86 and MAE = 7.5 months. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=86 SRC="FIGDIR/small/727196v1_ufig1.gif" ALT="Figure 1"> View larger version (20K): org.highwire.dtl.DTLVardef@11ead1org.highwire.dtl.DTLVardef@4f5c19org.highwire.dtl.DTLVardef@277de1org.highwire.dtl.DTLVardef@b95c9a_HPS_FORMAT_FIGEXP M_FIG C_FIG

MotivationThe tumor microenvironment (TME) dictates cancer progression and therapeutic response, yet translating TME subtypes into robust clinical biomarkers remains a significant challenge. Existing classification models typically rely on static gene signatures and cohort-dependent normalization, making them ill-suited for application to the small, unbalanced datasets common in early-phase clinical trials. To better guide drug development, methods are required that offer the flexibility to target specific biological contexts and bridge the gap between the discovery of tumor archetypes and their robust translation to individual patient samples. ResultsWe developed TumorArchetypeR, a modular R package that unifies unsupervised subtype discovery with the generation of rank-based, single-sample classifiers. By leveraging a systematic parameter grid search, the framework identifies stable, data-driven subtypes rather than relying on arbitrary defaults. Crucially, to ensure clinical translatability, the package includes a module to train a robust classifier using binary gene-pair rules, enabling prediction without cohort-level preprocessing. Applying TumorArchetypeR to colorectal cancer, we resolved the heterogeneity of fibrotic tumors, distinguishing an immunosuppressive "Immune-enriched/Fibrotic" state from an immune-excluded "Fibrotic/Myeloid" phenotype. Furthermore, we identified a distinct "Th/B-cell enriched" archetype associated with superior survival, a group largely obscured by existing pan-cancer models. With our rank-based classifier demonstrating robust performance on previously unseen samples, these findings highlight TumorArchetypeR as a scalable, end-to-end solution for refining patient stratification and optimizing precision oncology strategies. The TumorArchetypeR package and documentation are openly available on GitHub at https://github.com/lutgem/TumorArchetypeR.

Background: Safe reuse of multimodal hospital data for AI development is limited by the absence of reliable, context-aware deidentification across multimodal data and longitudinal patient data. Existing approaches are largely modality-specific and can indiscriminately remove clinically important information. Methods: We developed the Multimodal Anonymizer, a modular, locally deployable multi-agent framework integrating multimodal large language models, task-specific neural networks and rule-based transformations. We evaluated 16 orchestrator model configurations on a benchmark built from publicly available data and hospital data from our institution. The benchmark dataset included data from different origins: 250 MIMIC-IV patients with synthetically injected personally identifiable information (PII) supplemented with head CT, face images, handwriting, audio, German clinical-text datasets and local data. Primary outcomes were deidentification sensitivity and preservation of clinically important content; secondary analyses examined model characteristics, reproducibility, and performance against leading market and open-source solutions. Results: The best local configuration (the orchestrator being Qwen3-VL-235B-A22B-Thinking) achieved near-complete deidentification across all datasets, with per-patient sensitivity of 98.80% (95%-CI 97.20; 100), and per-PII sensitivity of 99.82% (95%-CI 99.76; 99.88). Critical clinical preservation was 99.60% (95%-CI 98.80; 100) per-patient, and clinical preservation was 99.61% (95%-CI 99.51; 99.71) per-file. All modalities achieved at least 98.30% sensitivity (lower bound 95%-CI). On our local data, the system achieved a deidentification sensitivity of 100% per-patient and per-PII; and a critical clinical preservation of 100% per-patient as well as a clinical preservation of 99.97% (95%-CI 99.91; 100) per-file. When comparing orchestrators, the leading local models were similar to proprietary models (GPT-5.2) in deidentification sensitivity while showing higher deidentification specificity. The Multimodal Anonymizer outperformed previous tools on most modalities. Conclusion: Near-complete, utility-preserving deidentification of multimodal clinical data is achievable with a unified, locally deployable multi-agent system, enabling safer large-scale reuse of hospital data for research and AI development.

Objectives: Patient-directed SMART on FHIR lets registries acquire longitudinal electronic health record data, but the payload requires substantial engineering before use. We present Registry Forge, an open-source pipeline that converts it into research-ready outputs. Materials and Methods: Registry Forge decodes and parses mixed C-CDA, HTML, RTF, PDF, and FHIR inputs, joins records to a canonical patient identifier, and emits a browser-viewable dashboard, an OMOP CDM v5.4 data set, GA4GH Phenopackets v2, a code inventory, and regex extractions of disease-specific narrative content. Results: Applied to the ALS Research Collaborative Study (94 participants, 56 US health systems), it processed 22,686 source files and 1,791 FHIR Bundles (109,599 resources); only 15.0% of files were full C-CDA. Discussion: This pipeline generalizes to any registry acquiring data through patient-directed SMART on FHIR. Conclusion: Registry Forge closes the acquisition-to-analysis gap with no server infrastructure and is openly available.