In modern oncology drug development, translational data science plays a critical role in connecting discovery biology with clinical development and patient care. For Olivier Harismendy, Vice President of Translational Data Science and Computational Biology at Zentalis Pharmaceuticals, the discipline sits at the intersection of translational biology, computational biology and medicine.
Its value lies in helping teams understand not only how a drug works, but also which patients are most likely to benefit from it. By integrating biological insight with computational approaches, translational data science can help de-risk development decisions, prioritize indications and generate evidence earlier in the clinical journey.
A key challenge is distinguishing between biomarkers that are scientifically interesting and those that are clinically actionable. Advanced tools such as spatial transcriptomics and spatial proteomics can reveal important biological signals, but they are not always practical for routine clinical use. A useful biomarker strategy must be compatible with real-world clinical workflows, including pathology lab capabilities, turnaround times and global accessibility. The goal is often to move from complex discovery platforms toward simpler, reliable surrogates that can be deployed in practice.
This is especially important in areas such as DNA damage response biology. At Zentalis, work around WEE1 inhibition highlights the complexity of targeting dynamic processes such as cell-cycle regulation. Unlike a static signaling pathway, cell-cycle biology involves timing, repair, transition points and more complex context-dependent vulnerabilities. Translational research is essential for understanding which cancer cells are most susceptible, how to widen the therapeutic window and how dosing schedules may spare normal cells while targeting tumors.
Another major challenge is connecting genomics, clinical outcomes, pathology and real-world data. Although each data source is powerful, they often remain siloed across teams, systems and departments. Harismendy emphasizes the importance of experts who can work across both forward and reverse translation: using preclinical data to interpret patient findings, and using patient data to generate new biological hypotheses.
AI and machine learning are also becoming increasingly useful, though their value depends on careful application. Large language models can help summarize internal knowledge, mine historical documents, support coding and accelerate routine work. However, they also require strong guardrails, particularly in regulated environments where data access, clinical trial blinding and accuracy are critical.
Looking ahead, oncology trial design is likely to become more data-driven and adaptive. External control arms based on real-world data may play a larger role, particularly in settings where single-arm studies need stronger contextual evidence. Liquid biopsy, cell-free DNA and molecular response measures may also become more important in early-phase development, helping teams assess response faster and in more convenient and cost-effective fashion compared toely traditional radiological assessment. As oncology development becomes more complex, translational data science will be central to making better decisions earlier. Its greatest promise lies in turning fragmented biological and clinical data into actionable evidence that can guide drug development, improve trial design and ultimately help match the right patients to the right therapies.
