James, Jose

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.

Introduction: Functional annotation of rare variants is key to the implementation of genomic medicine. We used molecular dynamics (MD) simulations to assess functional consequences of putative pathogenic variants (PVs) in the low-density lipoprotein receptor gene ( LDLR ). LDLR has a multidomain structure with seven LA repeats and local interactions between these repeats influence ligand binding. However, binding and recycling of the receptor-ligand complex require global changes in conformation. Using MD, we investigated the effects of PVs on long-range dynamics and LDL binding. Methods: An LA 4-6 repeat structure was created from a crystal structure of the extracellular domain (PDB: 1N7D). Structures were solvated with a water box containing TIP3P water molecules and simulated using the AMBER software package with the ff14SB forcefield. The wild type (WT) structure was first relaxed using MD; and subsequently PVs D245Y, F181Y and F181G were modeled. Dynamics were simulated for approximately one microsecond. Values for LDL binding for each PV were obtained from a prior study which used a radioisotope uptake assay. Results: WT domains demonstrated a stable conformation based on root mean square deviation. PVs disrupted the normal distribution in all domains significantly (p < 0.0001), broadening it and introducing multiple peaks. Projection along the top principal components (PCs) showed global destabilization with all PVs. The variant with the greatest impact on LDL binding, F181G, adopted a distinct alternate conformation. Comparison between PV and WT subspaces from PCA correlated with LDL binding (R 2 = 0.99, exponential regression). Structural analysis revealed that the LA5 calcium cage is involved in inter-domain salt bridges, which are disrupted in simulations of PVs. Conclusion: MD simulation of LA4-6 repeats in LDLR suggests PVs alter long-range conformational dynamics, and that such dynamics correlate with LDL binding. The impact of PVs on conformational dynamics and binding may be through disruption of interdomain salt bridges that are mediated by the acidic calcium cage. This study demonstrates the potential utility of MD simulations to functionally characterize LDLR variants.