SESSION K

Digital Pathology (DP) and multiplex immunohistochemistry (IHC) are rapidly transforming biomarker discovery, patient stratification and selection, and clinical trial efficacy assessment in the pharmaceutical industry. However, translating massive, high-dimensional image data into robust predictive models requires a deep understanding of both the underlying biology and the complex data processing pipelines. This tutorial is aimed to bridge the gap between tissue biology and data science. We will provide practical frameworks, pharmaceutical case studies, and actionable best practices for handling DP data—from the moment a slide is scanned to the final predictive model. Attendees will learn how to navigate the unique statistical challenges of spatial data, effectively manage pre-processing artifacts, and apply advanced machine learning and AI techniques to extract meaningful, actionable biomarker signatures.

This first part of the tutorial sets the stage by framing the evolution of tissue-based biomarkers. We will explore the transition from traditional, semi-quantitative IHC scoring by pathologists to modern, fully digitized Whole Slide Imaging (WSI) and multiplexing technologies.

The biological context of the data, current regulatory perspectives, and the future trajectory of DP in clinical development frameworks will be discussed.

The second part of the tutorial will focus on image processing and pre-data processing in DP.
The quality of downstream statistical analysis is entirely dependent on the upstream image processing. This session demystifies the “black box” of translating pixels into structured datasets.

Approaches and challenges in cell segmentation, tissue phenotyping, and spatial coordinate generation will be discussed. We will share practical frameworks and best practices for quality control (QC), handling artifacts, and ensuring high-fidelity data preparation.

The third part of the tutorial will provide the statistical frame on how to analyze DP data to build predictive biomarkers. It will cover study design, data quality evaluation, the core analytical challenge in digital pathology lies in translating highly complex, spatially resolved single-cell data into clinically actionable patient-level predictions. This session dives deep into the specialized statistical and machine learning frameworks required to navigate high-dimensional tissue datasets without falling into common overfitting traps. We will showcase real-world pharmaceutical examples of using AI applications and predictive modeling to identify predictive biomarker signatures for drug efficacy and patient response.

The key takeaways of this tutorial session:

• A clear conceptual framework of the end-to-end Digital Pathology pipeline.
• Best practices for quality control and pre-processing of spatial and image-derived tabular data.
• Concrete examples of applying machine learning and spatial statistics to DP data for clinical trial decision-making

Instructors’ Biography:

Talia Muram is currently Daiichi Sankyo’s executive director and head of translational pathology.

Zhaohua Lu is an Associate Director of Biostatistics at Daiichi Sankyo and a Ph.D.-trained statistician with more than 16 years of experience in statistical modeling, data science, and the design and implementation of clinical trials. Prior to joining Daiichi Sankyo, he served as a faculty member in the Department of Biostatistics at St. Jude Children’s Research Hospital and as a Principal Data Scientist at DiDi.

His research and professional expertise span the analysis of large-scale imaging, genomic, and natural language data, with particular interests in machine learning, Bayesian methods, statistical computing, and innovative clinical trial methodology. He has authored or co-authored more than 70 peer-reviewed clinical and methodological publications.

Li Liu is Head of the Precision Medicine Biostatistics Group at Daiichi Sankyo, leading statistical strategy and support for biomarker, translational science, and companion diagnostic (CDx) development activities. She led the statistical strategy and execution for the successful global regulatory submissions and approvals of quizartinib by the FDA, EMA, and PMDA. Prior to joining Daiichi Sankyo, Li held biostatistics positions at Sanofi and Pfizer, where she assumed roles of increasing responsibility across early- and late-phase oncology and hematology development programs. She has extensive experience in clinical trial design and execution, statistical strategies for translational science and CDx development, PK/PD modeling, and the application of innovative statistical methodologies to support precision medicine and oncology drug development.