Getting a new biomarker assay from algorithm to clinic is rarely a single-company problem. Most AI pathology companies can build the algorithm; few can also provide the analytical validation, global trial enrollment, and regulatory expertise to take it further.
CellCarta and Imagene AI expanded their partnership to coordinate AI-powered biomarker and companion diagnostic (CDx) programs from development to regulatory readiness. Two programs with unnamed global pharmaceutical companies are already active. One focuses on immunohistochemistry (IHC)-based CDx development; the other applies Imagene AI's Lung Prediction Panel in non-small cell lung cancer (NSCLC).
"Pharma companies need a practical path for translating AI-powered biomarker and image-analysis capabilities into assays that can support clinical development and companion diagnostic programs," said Michael Hreczuck, senior vice president of business development at Imagene AI.
AI Pathology: From IHC Quantification to H&E-Based Prediction
The collaboration covers four areas of the biopharma development pipeline:
- Assay validation: fit-for-purpose analytical testing for regulatory submission
- Clinical deployment: global trial infrastructure for sponsored biomarker programs
- Regulatory support: CDx program strategy from development to submission
- Program coordination: single-partner model spanning assay design to regulatory readiness
"Bringing AI-powered approaches into global drug development requires more than a strong algorithm—it requires the ability to validate, deploy and scale within the rigor of clinical development," said Christopher Ung, chief scientific business officer at CellCarta.
The three approaches differ in staining requirements, scoring precision, and regulatory status:
| Approach | Tissue staining | Scoring method | Turnaround | Regulatory status |
|---|---|---|---|---|
| Conventional IHC | Dedicated stain | Semiquantitative, observer-dependent | Days–weeks | CDx-eligible |
| AI-powered IHC (Oncology Intelligence CDx) | Dedicated stain | Continuous, subcellular, spatial | Hours | In development |
| H&E-based prediction (Lung Prediction Panel) | H&E only | Model output; no additional stain | Minutes | RUO |
What Does the Research Show About H&E-Based Biomarker Prediction?
The scientific case for hematoxylin and eosin (H&E)-based biomarker prediction has grown substantially. A 2019 study in the Journal of Pathology Informatics showed that a multi-field-of-view deep learning model could predict programmed death-ligand 1 status in NSCLC directly from H&E whole-slide images (WSIs). A July 2025 Nature Medicine study demonstrated real-world clinical utility for an H&E-based EGFR (epidermal growth factor receptor) biomarker, achieving an area under the curve of 0.890 in a prospective silent trial on primary samples.
AI Pathology and CDx Validation: Converging Toward Clinical Use
Few companies can take an algorithm from assay design through validated CDx deployment within a single relationship. This collaboration addresses a gap that has limited AI pathology adoption in biopharma. Growing research on H&E-based biomarker prediction adds scientific weight to the technical approach at the center of the collaboration's active NSCLC program.
This article is based on a press release issued by CellCarta and was produced under The Scientist's Editorial Policies.
