AI agent applications in digital traceability for cell and gene therapy: COI(Chain of Identity) and coc(Chain of Custody) management

October 15–16, 2025 | Philadelphia, USA The American Society of Gene & Cell Therapy (ASGCT) will host the Advancing Cell + Gene Therapies for Cancer meeting, dedicated to breakthroughs and challenges in oncology-focused CGT. Topics include tumor heterogeneity, viral vector & cell therapy combinations, mechanisms of resistance, gene editing optimization, and cross-platform integration.

Bio LIMS INC is proud to share that our work “AI agent applications in digital traceability for cell and gene therapy: COI / COC management” has been peer-reviewed and accepted as an official abstract at this meeting. 👉 The abstract can be downloaded directly from the Advancing Cell + Gene Therapies for Cancer | ASGCT under the Poster Presentations section.

Introduction

Cell and Gene Therapy (CGT) represents a revolutionary frontier in precision medicine, offering transformative treatments for previously untreatable conditions through direct modification of genetic material or introduction of functional cells. However, the highly personalized, customized nature of CGT products presents unprecedented challenges for end-to-end traceability throughout the product lifecycle. Unlike traditional pharmaceuticals, each CGT product is essentially a "one-time" bespoke therapy for a specific patient, requiring precise Chain of Identity (COI) from donor to recipient and rigorous Chain of Custody (COC) documentation. Current traceability systems struggle with data heterogeneity across multiple stakeholders, complex multi-step manufacturing processes, stringent time sensitivity requirements, diverse regulatory frameworks, and critical privacy concerns regarding sensitive genetic information.

Methods

This research develops an innovative AI Agent-based digital traceability framework specifically designed for the unique requirements of CGT products. The system architecture features a distributed network of specialized Agents: donor information Agent for donor screening and risk assessment; manufacturing process Agent for real-time monitoring of complex bioproduction steps; quality control Agent for automated analysis of multi-dimensional quality metrics; logistics tracking Agent for intelligent cold chain management; and clinical application Agent for patient-product association and long-term safety monitoring. Each Agent leverages Large Language Model (LLM) technology to process multimodal data (text, images, sensor readings), perform semantic understanding, and execute domain-specific reasoning. The system integrates knowledge graph technology for unified data representation. Implementation follows a microservices architecture with containerized deployment, multi-model database design, and standardized API interfaces to ensure scalability and interoperability.

Results

The system was validated across five representative CGT organizations (including CAR-T companies, stem cell therapy centers, and gene editing therapy firms) with many real treatment cases from January 2022 to June 2024. Performance metrics demonstrate significant improvements over traditional systems:

Key info extraction accuracy

96.7%

vs. 78.4% baseline

Query response time

3.0 s

−72.2% vs. 10.8 s

Anomaly detection F1

0.95

+55.7% vs. 0.61

Manual intervention

17%

−83% vs. 100%

Data consistency

94.5%

+26.2% vs. 68.3%

In practical applications, the system successfully managed a cold chain interruption incident involving a CAR-T product, preventing potential product loss while ensuring patient safety.

Conclusions

This research demonstrates that AI Agent technology can transform CGT traceability from passive data recording to active intelligence, addressing critical industry challenges through cognitive capabilities, autonomous decision-making, and cross-domain collaboration. The system successfully bridges the gap between complex CGT workflows and regulatory requirements while maintaining stringent privacy protections. Future development will focus on integrating decentralized identity (DID) technology, digital twin capabilities, and quantum-resistant cryptography to further enhance system robustness. The framework has broader applicability beyond CGT to mRNA vaccines, organoids, and synthetic biology, positioning AI Agents as foundational digital infrastructure for the emerging bioeconomy. As AI and biotechnology continue to converge, this research provides a replicable paradigm for lifecycle management of high-value biological products, supporting the evolution of global healthcare toward autonomy, trustworthiness, and intelligence.