Dr. Diego Chowell is an Assistant Professor in the Departments of Immunology and Immunotherapy, Oncological Sciences, and Artificial Intelligence and Human Health at the Icahn School of Medicine at Mount Sinai, New York.

Routine blood tests and basic clinical data, collected for nearly every cancer patient worldwide, represent an underexplored resource for understanding systemic immune status and predicting response to therapy. These measurements capture critical aspects of inflammation, immune cell balance, and overall physiological condition, offering a cost-effective and scalable complement to tumor genomics and tissue-based biomarkers.
In this seminar, Dr. Diego Chowell will introduce SCORPIO, a machine learning system that leverages these widely available data to predict outcomes for immune checkpoint inhibitor (ICI) therapy across multiple cancer types. Validated in real-world settings and 10 phase 3 clinical trials, SCORPIO outperforms FDA-approved biomarkers and more complex machine learning models, highlighting the untapped potential of everyday clinical data to advance precision immuno-oncology.

Dr. Charles Weijer from Western University (Canada) explored the evolving ethical landscape of modern randomized controlled trials and their implications for research practice. As Professor of Medicine (Critical Care) and Philosophy, Dr. Weijer brought unique expertise from his leadership in developing the first international ethics guidelines for cluster randomized trials and his work on WHO guidance for health policy research.

Attendees learned how pragmatic RCTs, designed to inform real-world clinical decisions, challenge traditional informed consent approaches when conducted across diverse patient populations and complex healthcare settings. Dr. Weijer examined why cluster RCTs, which randomize groups rather than individuals, require rethinking research ethics beyond individual participant protection, and how these challenges offer opportunities to deepen our understanding of core ethical principles in clinical research.

Dr. Nikolaus Schultz from Memorial Sloan Kettering Cancer Center (USA) will present MSK-CHORD, a comprehensive clinicogenomic dataset that combines natural language processing with structured medical data from nearly 25,000 cancer patients.

Learn how machine learning models trained on MSK-CHORD outperform models based solely on genomic data or cancer staging, and discover findings from over 700,000 annotated radiology reports, including a relationship between SETD2 mutation and reduced metastatic potential in immunotherapy-treated lung adenocarcinoma.

DOI: https://doi.org/10.1038/s41586-024-08167-5

Dr. Ravi B. Parikh from Emory University School of Medicine (USA) presented on “Real-world Emulation of Clinical Trials: Novel methodologies & Applications for Trial Design and Clinical Practice”. Dr. Parikh presented his TrialTranslator framework, which uses machine learning to evaluate how cancer treatment trials generalize to real-world patients. Dr. Parikh shared compelling insights about the gap between clinical trial results and actual patient outcomes.

The post-presentation discussion was also valuable, with Dr. Qi Long helping to address technical methodological questions.

DOI: https://doi.org/10.1038/s41591-024-03352-5

Dr. Anna Dorothea Wagner from Lausanne University Hospital (Switzerland) presented on "Gender Medicine & Oncology” highlighting how biological sex and gender constructs influence health outcomes across medical care. She explained that biological sex determines crucial host factors affecting cancer susceptibility and medication responses, while demonstrating evidence of “sexual dimorphism” in cancer – showing distinct biological differences when cancers occur in male versus female patients, even in non-reproductive organs. Wagner emphasized that gender shapes lifestyle choices and risk factor exposure, with continuous epigenetic interactions between biological sex and gender. She also addressed how clinical trials typically analyze sex/gender differences only for primary endpoints, with observed differences rarely impacting drug registration or treatment guidelines.

DOI: 10.1056/NEJMp2403550