A Cloud-Enabled LIMS-QMS Platform With Large Language Model-Assisted Standard Operating Procedure Generation for...

Theranostic development centers operate under stringent good laboratory practice (GLP) and good manufacturing practice (GMP) requirements that fragmented, paper-based workflows cannot adequately support. This paper presents the design, regulatory compliance mapping, and pilot deployment of the Laboratory Tracking Software (LTS) platform, a cloud-ready, web-based system that unifies laboratory information management system (LIMS) and quality management system (QMS) functions under a risk-based Computer Software Assurance (CSA) framework aligned with U.S. Food and Drug Administration (FDA) 21 CFR Part 11 and ISO/IEC 17025. The platform operationalizes a 6×6 deterministic request-and-service lifecycle across nine governance modules and embeds a large language model (LLM)-assisted module that, from a single service-name input, generates structured standard operating procedures (SOPs), identifies required consumables and instruments, performs operational risk assessment, and proposes an executable workflow. Pilot deployment across 187 laboratory requests is reported descriptively (23-day mean turnaround time, 1,456 quality-control checkpoints, and 43 deviation-handling episodes); because no comparable pre-deployment dataset could be reconstructed, these values are not interpreted as efficiency or error-reduction gains. A 21 CFR Part 11-aligned control matrix and a CSA-based verification package document authenticated electronic approvals, activity records, two-factor authentication (2FA), and role-based access control (RBAC); a consolidated high-risk user-acceptance campaign executed 31 of 31 scripted scenarios successfully, with independent reviewer and approver sign-off still pending. Beyond the original single BRAF droplet digital PCR (ddPCR) illustrative case, a document-level evaluation was performed across a locked corpus of 20 laboratory services (10 analysis and 10 synthesis workflows). All 280 mandatory template elements were present (100.0% template adherence), and none of the 80 predefined automated document-control checks was flagged (0.0%). Expert-adjudicated error rate, human edit distance, and time saving were not measured and are deferred to a prospective follow-up study. Comparative analysis against QT9 ERP/QMS, RAMS®, and SENAITE LIMS positions the platform as a governance-first, theranostic-specific informatics system with native LLM-assisted documentation generation, a capability absent from existing commercial and open-source alternatives. The results indicate that integrating LLM-assisted document generation into a regulated laboratory informatics platform is technically feasible; however, given the hallucination risk documented in the medical LLM literature, systematic human-in-the-loop validation of generated content remains a prerequisite for unsupervised deployment in GLP/GMP settings. The revised design description additionally specifies a template-constrained prompting strategy, a generation-record schema supporting prompt and output reconstruction, a model-change-control procedure, an explicit LLM data-privacy boundary, and alignment with the January 2026 FDA/EMA guiding principles of good artificial intelligence (AI) practice in drug development.