Lunit, Agilent Partner on AI-Based Companion Diagnostics to Speed Cancer Drug Development

Lunit and Agilent Team Up to Build AI Companion Diagnostics for Faster Oncology Drug Development

AI company Lunit announced a collaboration with Agilent Technologies to co-develop AI-based companion diagnostics (CDx) for new drug development. The focus: combining Lunit's AI pathology analysis with Agilent's tissue-based diagnostic capabilities to accelerate biomarker-driven trials and approvals.

Companion diagnostics help determine which patients are likely to benefit from a specific therapy or face higher risk of side effects. Agilent, a major provider of analytical solutions used across pharma and biotech, reported $6.5 billion in revenue last year.

What's launching first

Early collaboration will target AI solutions used inside clinical trials-think biomarker analysis that informs patient selection and response prediction. Mid to long term, the companies plan to seek CDx approvals aligned with global oncology programs.

"If Agilent's global diagnostics platform is combined with Lunit's validated AI pathology analysis technology, pharmaceutical companies will be able to bring new drugs to market far more accurately and quickly than before," said Seo Beom-seok, Lunit's CEO. Nina Green, who leads Agilent's clinical diagnostics business, noted the partnership aims to drive wider adoption of precision medicine through AI-enabled CDx.

Why this matters for IT and development teams

• Data pipelines: Expect whole-slide images (WSI), H&E/IHC data, and clinical endpoints-requiring scalable storage, performant tiling/patching, and standardized metadata.
• MLOps in regulated settings: Versioned datasets, traceable experiments, and locked model artifacts for Good Clinical Practice (GCP) and audit-readiness.
• Validation at scale: Site-to-site variability, scanner differences, staining drift-require robust cross-site validation and continuous monitoring.
• Interoperability: Plan for DICOM-WSI, LIMS/EDC integration, and potential FHIR-based data exchange to fit into sponsor and CRO systems.
• Security and privacy: PHI handling, role-based access, encryption at rest/in transit, and compliance with HIPAA/GDPR where applicable.
• Human-in-the-loop: Pathologist review workflows, uncertainty thresholds, and explainability artifacts to support clinical decisions and submissions.
• Regulatory documentation: SaMD classification, intended use statements, predicate analysis (if any), and 21 CFR Part 11-compliant audit trails.

Practical technical scope to anticipate

• Biomarker scoring from tissue (e.g., quantifying IHC signals, morphology-driven features) and patient stratification logic for trials.
• Model robustness: stain normalization, scanner harmonization, artifact detection, and domain adaptation.
• Deployment models: on-prem or hybrid due to data residency; containerized inference with GPU scheduling; offline-capable edge nodes in hospitals.
• Observability: per-site performance dashboards, drift metrics, and automated alerts tied to retraining triggers.
• Integration: API/SDK endpoints for CRO tools, LIMS, and trial randomization systems with strict SLAs.

If you're building in this space

• Data readiness: secure data agreements, de-identification pipelines, and high-quality annotations with adjudication.
• Quality systems: implement QMS early (SOPs, CAPA, change control) to avoid rework before pivotal studies.
• Documentation: keep submission-ready artifacts-risk management files, verification/validation reports, and usability studies.
• Team skills: blend ML, digital pathology, MLOps, and regulatory engineering to meet trial timelines.

For background on how companion diagnostics are regulated, see the FDA overview: FDA: Companion Diagnostics. Learn more about Agilent's diagnostics and life sciences platform here: Agilent Technologies.

If you're upskilling for AI in healthcare and regulated ML, explore role-based learning paths: Complete AI Training - Courses by Job.