FMRP-LEAN: A HIPAA-Compliant AI-Augmented LIMS Architecture for End-to-End Clinical Assay Workflow Optimization

Clinical biomarker workflows, such as the quantification of Fragile X Messenger Ribonucleoprotein (FMRP), are often hindered by fragmented, spreadsheet-based tracking and manual quality control (QC) processes. These legacy methods introduce operational risks, delayed reporting, and poor visibility into sample status. In regulated healthcare environments, these challenges are compounded by the need to maintain strict HIPAA compliance, ensure data residency, and safely integrate AI tools without compromising patient privacy or clinical accuracy.

FMRP-LEAN addresses these inefficiencies by formalizing the biospecimen lifecycle through a finite-state machine. Each sample is assigned a unique identifier (UUIDv7) and transitions through explicit, auditable states—from collection to reporting. The system is built on a self-hosted Supabase/PostgreSQL stack, utilizing a hybrid edge-internal security topology. This design ensures that all protected health information (PHI) remains within hospital-controlled infrastructure, while a public edge host manages external connectivity via encrypted tunneling and loopback-only service binding.

Unlike many healthcare AI systems that process patient-level data, FMRP-LEAN employs a "governance-constrained" AI module. The system generates operational briefings and plate-planning suggestions based exclusively on aggregate projections (e.g., total pending assays or mean dwell times). By enforcing a formal PHI isolation constraint, the architecture ensures that no patient identifiers are ever included in AI prompts. Furthermore, the system includes a deterministic fallback: if the AI service is unavailable, the system defaults to displaying raw structured aggregates, ensuring that operational continuity is never dependent on the AI model.