Citations

Abstract Randomized controlled trials (RCTs) provide high internal validity but often rely on restrictive eligibility criteria that limit generalizability and complicate real-world trial emulation. We propose AERO (AI Agent for Adaptive Eligibility Refinement and Optimization), an agentic framework that systematically adapts clinical trial eligibility criteria for application to electronic health record data. AERO integrates external clinical knowledge sources and large language model–based reasoning to classify criteria as strict inclusion, safety exclusion, confounder, or operational artifact. We evaluated AERO by emulating the WARCEF trial using Mayo Clinic Platform data restricted to the pre-trial completion period. Emulation with optimized criteria yielded a hazard ratio of 1.561 (p = 0.0605), consistent with the original neutral trial finding (HR = 1.01, p = 0.91). An ablation analysis demonstrated that eligibility handling decisions materially influence observed treatment effects. These results highlight the importance of systematic, knowledge-informed eligibility refinement in real-world evidence generation.

Abstract Background Electronic health record (EHR)–based prognostic modeling is increasingly used in oncology, yet incorporating pharmacogenomic (PGx) knowledge derived from experimental systems into clinical prediction frameworks remains challenging. This gap is driven by fundamental mismatches between controlled drug–mutation assays and heterogeneous, incomplete real-world clinical data. Methods We propose a representation transfer framework that integrates PGx embeddings learned from large-scale in vitro pharmacogenomic screens into patient-level EHR models. A frozen pharmacogenomic encoder is used to generate interaction-aware embeddings from patient mutation profiles and administered therapies, which are aggregated into a fixed-length PGx Complementarity Representation. These representations are incorporated into multimodal survival prediction models alongside standard clinical features. Performance was evaluated using systematic modality ablation analyses, attribution analyses, and exploratory unsupervised representation analyses. Results Integrating PGx embeddings yielded consistent performance improvements across all evaluated modality combinations. Relative gains were largest in modality-sparse settings, where baseline EHR features encode limited biological context, and were attenuated—but remained significant—in biologically enriched configurations. Attribution analyses indicated that PGx embeddings contributed non-redundant predictive signal beyond standard clinical features. Exploratory unsupervised analyses further demonstrated that the learned representations exhibit interpretable association patterns aligned with known therapeutic exposures and pathway-level associations. Conclusion These findings suggest that externally learned pharmacogenomic representations can be transferred into real-world EHR models as a context-dependent, non-redundant augmentation. By framing PGx knowledge as an interaction-aware representation rather than a mechanistic model, this work provides an informatics framework for integrating experimental pharmacogenomic data into clinical prediction tasks in a reproducible and interpretable manner.

10556 Background: The coexistence of interstitial lung diseases (ILD) and lung cancer (LC) presents a clinical challenge. ILD, including idiopathic pulmonary fibrosis (IPF) and non-IPF entities, is associated with an elevated risk of LC. However, the clinical features, treatment patterns, and survival outcomes of ILD-LC, especially IPF-LC, remain underexplored. We aimed to comprehensively evaluate survival outcomes and risk factors for mortality of ILD-LC patients based on a real-world long-term clinical cohort. Methods: A total of 1,341 eligible patients for this study were drawn from the Epidemiology and Genetics of Lung Cancer research program database (Mayo Clinic) including 96 IPF-LC, 111 Non-IPF ILD-LC, 99 ILD-only, and 1,035 propensity-matched LC-only patients (Controls). Data were collected on demographics, comorbidities, tumor features, treatment, and patient outcomes. Survival analyses used Kaplan-Meier methods; risk factors were assessed via Cox Proportional Hazards models. All comparisons reported below were statistically significant with p-values≤0.001 unless specified. Results: Among the 20,470 primary LC patients who were diagnosed from 1997 through 2016 and followed to 2021, 207 (1.0%) had ILD-LC, of which 96 (46.4%) had IPF. Comparing to Controls, ILD-LC patients were 4 years older at the time of LC diagnosis, 63.8% (vs. 52.8%) male, had a higher rate of former smokers (60.4% vs. 47.3%), a lower proportion of adenocarcinoma (37.7% vs. 57.5%), and a higher proportion of squamous cell (28.0% vs. 20.3%) and small cell carcinoma (13.5% vs. 8.4%). Although no significant differences in either LC stage or grade, ILD-LC patients had a significantly higher mortality rate (90.3% vs. 78.8%), particularly among IPF-LC patients (97.9% vs. 83.8%). ILD-LC patients were less likely to undergo surgery or receiving pharmacological treatments (73.9% vs. 93.5%). The median 5-year overall survival rate (OS, months) were 49.5 for Controls, 35.2 for Non-IPF ILD-LC patients, and 15.9 for IPF-LC patients. Analysis of six most common comorbidities revealed that pulmonary hypertension was associated with worse OS in ILD-LC patients compared to those without it (11.9% vs. 20.1%, P < 0.05); however unexpectedly, hypertension in ILD-LC patients was associated with better OS (21.3% vs. 14.4%) and emphysema in IPF-LC patients with better OS (7.9% vs. 3.4%, P < 0.05) when compared to those without the two comorbidities, respectively. Conclusions: This study highlights the poorer survival outcomes in ILD-LC patients, particularly IPF-LC patients, emphasizing the need for further investigation into optimized therapies and comorbidity management. Our findings also call for attention to two unexpected associations of hypertension and emphysema, warranting in-depth research on the possible biological and physiological mechanisms.

Abstract Objective The fast accumulation of vast pharmacogenomics data of cancer cell lines provide unprecedented opportunities for drug sensitivity prediction (DSP), a crucial prerequisite for the advancement of precision oncology. Recently, Generative Large Language Models (LLM) have demonstrated performance and generalization prowess across diverse tasks in the field of natural language processing (NLP). However, the structured format of the pharmacogenomics data poses challenge for the utility of LLM in DSP. Therefore, the objective of this study is multi-fold: to adapt prompt engineering for structured pharmacogenomics data toward optimizing LLM’s DSP performance, to evaluate LLM’s generalization in real-world DSP scenarios, and to compare LLM’s DSP performance against that of state-of-the-science baselines. Methods We systematically investigated the capability of the Generative Pre-trained Transformer (GPT) as a DSP model on four publicly available benchmark pharmacogenomics datasets, which are stratified by five cancer tissue types of cell lines and encompass both oncology and non-oncology drugs. Essentially, the predictive landscape of GPT is assessed for effectiveness on the DSP task via four learning paradigms: zero-shot learning, few-shot learning, fine-tuning and clustering pretrained embeddings. To facilitate GPT in seamlessly processing the structured pharmacogenomics data, domain-specific novel prompt engineering is employed by implementing three prompt templates (i.e., Instruction, Instruction-Prefix, Cloze) and integrating pharmacogenomics-related features into the prompt. We validated GPT’s performance in diverse real-world DSP scenarios: cross-tissue generalization, blind tests, and analyses of drug-pathway associations and top sensitive/resistant cell lines. Furthermore, we conducted a comparative evaluation of GPT against multiple Transformer-based pretrained models and existing DSP baselines. Results Extensive experiments on the pharmacogenomics datasets across the five tissue cohorts demonstrate that fine-tuning GPT yields the best DSP performance (28% F1 increase, p-value= 0.0003) followed by clustering pretrained GPT embeddings (26% F1 increase, p-value= 0.0005), outperforming GPT in-context learning (i.e., few-shot). However, GPT in the zero-shot setting had a big F1 gap, resulting in the worst performance. Within the scope of prompt engineering, performance enhancement was achieved by directly instructing GPT about the DSP task and resorting to a concise context format (i.e., instruction-prefix), leading to F1 performance gain of 22% (p-value=0.02); while incorporation of drug-cell line prompt context derived from genomics and/or molecular features further boosted F1 score by 2%. Compared to state-of-the-science DSP baselines, GPT significantly asserted superior mean F1 performance (16% gain, p-value<0.05) on the GDSC dataset. In the crosstissue analysis, GPT showcased comparable generalizability to the within-tissue performances on the GDSC and PRISM datasets, while statistically significant F1 performance improvements on the CCLE (8%, p-value=0.001) and DrugComb (19%, p-value=0.009) datasets. Evaluation on the challenging blind tests suggests GPT’s competitiveness on the CCLE and DrugComb datasets compared to random splitting. Furthermore, analyses of the drug-pathway associations and log probabilities provided valuable insights that align with previous DSP findings. Conclusion The diverse experiment setups and in-depth analysis underscore the importance of generative LLM, such as GPT, as a viable in silico approach to guide precision oncology. Availability https://github.com/bioIKEA/SensitiveCancerGPT

Abstract Objectives Heart failure (HF) impacts millions of patients worldwide, yet the variability in treatment responses remains a major challenge for healthcare professionals. The current treatment strategies, largely derived from population based evidence, often fail to consider the unique characteristics of individual patients, resulting in suboptimal outcomes. This study aims to develop computational models that are patient-specific in predicting treatment outcomes, by utilizing a large Electronic Health Records (EHR) database. The goal is to improve drug response predictions by identifying specific HF patient subgroups that are likely to benefit from existing HF medications. Materials and Methods A novel, graph-based model capable of predicting treatment responses, combining Graph Neural Network and Transformer was developed. This method differs from conventional approaches by transforming a patient's EHR data into a graph structure. By defining patient subgroups based on this representation via K-Means Clustering, we were able to enhance the performance of drug response predictions. Results Leveraging EHR data from 11 627 Mayo Clinic HF patients, our model significantly outperformed traditional models in predicting drug response using NT-proBNP as a HF biomarker across five medication categories (best RMSE of 0.0043). Four distinct patient subgroups were identified with differential characteristics and outcomes, demonstrating superior predictive capabilities over existing HF subtypes (best mean RMSE of 0.0032). Discussion These results highlight the power of graph-based modeling of EHR in improving HF treatment strategies. The stratification of patients sheds light on particular patient segments that could benefit more significantly from tailored response predictions. Conclusions Longitudinal EHR data have the potential to enhance personalized prognostic predictions through the application of graph-based AI techniques.

Abstract Background Error analysis plays a crucial role in clinical concept extraction, a fundamental subtask within clinical natural language processing (NLP). The process typically involves a manual review of error types, such as contextual and linguistic factors contributing to their occurrence, and the identification of underlying causes to refine the NLP model and improve its performance. Conducting error analysis can be complex, requiring a combination of NLP expertise and domain-specific knowledge. Due to the high heterogeneity of electronic health record (EHR) settings across different institutions, challenges may arise when attempting to standardize and reproduce the error analysis process. Objectives This study aims to facilitate a collaborative effort to establish common definitions and taxonomies for capturing diverse error types, fostering community consensus on error analysis for clinical concept extraction tasks. Materials and Methods We iteratively developed and evaluated an error taxonomy based on existing literature, standards, real-world data, multisite case evaluations, and community feedback. The finalized taxonomy was released in both .dtd and .owl formats at the Open Health Natural Language Processing Consortium. The taxonomy is compatible with several different open-source annotation tools, including MAE, Brat, and MedTator. Results The resulting error taxonomy comprises 43 distinct error classes, organized into 6 error dimensions and 4 properties, including model type (symbolic and statistical machine learning), evaluation subject (model and human), evaluation level (patient, document, sentence, and concept), and annotation examples. Internal and external evaluations revealed strong variations in error types across methodological approaches, tasks, and EHR settings. Key points emerged from community feedback, including the need to enhancing clarity, generalizability, and usability of the taxonomy, along with dissemination strategies. Conclusion The proposed taxonomy can facilitate the acceleration and standardization of the error analysis process in multi-site settings, thus improving the provenance, interpretability, and portability of NLP models. Future researchers could explore the potential direction of developing automated or semi-automated methods to assist in the classification and standardization of error analysis.