Lauri Diehl, Executive Director of Research Pathology at Gilead Sciences, spoke ahead of Biomarkers US in San Diego about the growing role of pathology in biomarker discovery, the transformative impact of digital pathology and artificial intelligence, and how pathology-enabled precision medicine may evolve over the next five years.
Where does pathology add the most value in turning biological signals into decision-ready evidence?
Diehl described pathology as a discipline that influences every stage of drug development, from the earliest research concepts through to post-approval questions surrounding marketed therapies.
She identified two areas where pathology delivers particularly significant value.
The first is patient stratification. Understanding target expression within tumours—and determining whether those same targets are present in healthy tissues where they could contribute to safety concerns—is critical for identifying the patients most likely to benefit from treatment. This information can directly inform pivotal trial design and patient selection strategies.
The second is tissue pharmacodynamics. Over the past four to five years, the industry has increasingly moved beyond relying exclusively on peripheral blood biomarkers to assess drug activity.
Diehl's team focuses on identifying tissue-based biomarkers that confirm target engagement and expected biological responses directly within tumour tissue. This information is especially valuable during early dose-escalation studies, where investigators need to determine whether sufficient drug activity has been achieved at the intended site of action.
How has the role of pathology changed in biomarker discovery and translational medicine?
Diehl has witnessed substantial evolution in the field since the early development of tissue biomarkers in oncology.
She pointed to the experience surrounding Herceptin as a defining moment that demonstrated both the promise and limitations of manually scored biomarkers. Human assessment of tissue samples is inherently variable, and much of the industry's effort over the last two decades has focused on standardising scoring approaches.
Despite these efforts, variability remains a challenge whenever biomarker data is generated outside highly centralised laboratory environments.
The most significant change today is the rapid transition toward AI-assisted and automated biomarker scoring.
Diehl emphasised that this shift should not be viewed as a replacement for pathologists. Instead, she sees pathologists moving away from repetitive execution tasks and toward roles centred on scientific interpretation, quality oversight, and expert decision-making.
The expected outcome is cleaner, more consistent biomarker data that supports both internal drug development decisions and regulatory confidence in biomarker-driven programmes.
What role do digital pathology and AI now play in quantifying tissue biology more consistently or at greater scale?
Diehl described several ways in which digital pathology is already transforming practice.
In diagnostic pathology, AI-assisted review can improve efficiency and accuracy while keeping pathologists firmly involved in oversight and final interpretation.
Within drug development, the most relevant applications involve quantifying target expression for large-molecule therapeutics, including:
• Antibody-drug conjugates (ADCs)
• T-cell engagers
• Related biologic modalities
Traditional scoring approaches, such as H-scores, provide only semi-quantitative measurements and can be affected by substantial inter-observer variability.
Digital pathology converts visual observations into numerical data that can be analysed more objectively and statistically. This creates opportunities for greater reproducibility and stronger scientific conclusions.
Diehl also highlighted a more emerging and controversial area: AI-generated tissue biomarkers derived from models that are not fully explainable, often referred to as "black-box" models.
Her perspective is pragmatic. The clearest near-term benefits will come from automating processes that are already well understood and removing sources of human variability. Less interpretable AI models remain promising but largely unproven, and the coming years will determine their true value.
What are the biggest challenges in standardising pathology-derived biomarker data across studies, platforms, and sites?
According to Diehl, the primary historical obstacle has been infrastructure.
For many years, digital pathology systems were difficult to deploy at the scale required for routine clinical use outside central laboratories.
That situation has changed considerably.
Several commercial platforms are now widely available and increasingly accepted by the pathology community. Diehl noted that Europe has generally moved ahead of the United States in adopting digital pathology technologies.
At the same time, regulatory agencies have become more engaged with the field and have worked to provide greater clarity regarding implementation and validation requirements.
The remaining challenge is demonstrating value.
Several therapies currently in development will effectively test whether digital pathology-derived biomarker data can meaningfully improve patient outcomes and development decisions.
Diehl expressed confidence that cleaner and more reproducible data will ultimately justify adoption, both from a scientific perspective and in meeting increasingly rigorous regulatory expectations.
Looking five years ahead, what would success look like for pathology-enabled precision medicine?
Diehl suggested that the answer depends on the stakeholder being considered.
For clinicians and patients, success would mean faster and more accurate information to support treatment decisions, particularly in oncology. Delivering better-informed therapeutic choices alone would represent a meaningful achievement.
For pharmaceutical companies, success is somewhat broader and more complex.
She suggested that consistently meeting regulatory expectations for biomarker quality and reproducibility would be an important milestone, even before any individual clinical trial demonstrates a direct impact on outcomes.
In the near term, achieving more consistent and reliable biomarker data may itself be considered a significant success measure.
While the full potential of pathology-enabled precision medicine is still emerging, Diehl believes that improving data quality and reducing variability will provide the foundation upon which future advances are built.
KEY TAKEAWAYS
• Pathology contributes across the entire drug development continuum, with particular value in patient stratification and tissue pharmacodynamics.
• Tissue-based biomarkers are becoming increasingly important for confirming target engagement and biological activity directly within tumours.
• AI-assisted pathology is shifting the role of pathologists from manual scoring toward expert interpretation and scientific oversight.
• Digital pathology enables the conversion of visual tissue observations into objective, quantitative datasets suitable for statistical analysis.
• Infrastructure and regulatory acceptance have improved substantially, reducing historical barriers to digital pathology adoption.
• The immediate value of digital pathology may be improved consistency and reproducibility of biomarker data, even before measurable impacts on clinical outcomes are fully demonstrated.
• Over the next five years, pathology-enabled precision medicine is expected to support faster, more reliable decision-making for both clinicians and drug developers.
