Researchers across pharma and academia are reducing oncology drug timelines with Tempus Lens
Scientific leads from academia, biotech, and pharma share how an AI-focused platform with multimodal real-world data is helping them answer critical questions in months, not years.
Leveraging complex, multimodal real-world data (RWD) is table-stakes in oncology research today. However, integrating genomics, transcriptomics, clinical data, and imaging at scale still presents challenges. To explore how researchers are overcoming these hurdles, Tempus recently hosted a webinar with leaders from across the life sciences ecosystem.
The discussion, moderated by Katie Igartua, PhD, VP of Translational Research at Tempus, featured a panel of experts who use the Tempus Lens platform to advance their work. They shared how an integrated, AI-enabled workbench helps them answer critical questions faster—from characterizing rare cancers to understanding treatment landscapes and evaluating novel biomarkers. Panelists included Olivier Harismendy, PhD, VP of Translational Data Science at Zentalis Pharmaceuticals; Xiaolin Zhu, MD, PhD, a genitourinary medical oncologist and physician-scientist at the University of California, San Francisco; Wilson Ngai, MSc, PharmD, MBA, Senior Medical Director and US GI Oncology Medical Lead at Eisai; and Shelley MacNeil, PhD, Director of Data Solutions at Tempus.
How are you using multimodal RWD to address specific research challenges in your field?
Olivier Harismendy, PhD: At Zentalis, we are developing a WEE1 inhibitor for platinum-resistant ovarian cancer. Our companion diagnostic is an IHC assay for cyclin E1 protein expression, which defines a new patient population that is not yet well-characterized. It was critical for us to understand the prognostic value of this biomarker and how these patients perform on the current standard of care. Since protein expression data isn't available for an unapproved assay, we used the extensive transcriptomic data in Lens to use CCNE1 gene expression as a surrogate. The platform's flexibility allowed us to iteratively explore hypotheses about treatment sequencing and outcomes in this novel subgroup.
Xiaolin Zhu, MD, PhD: I focus on rare cancers and complex, treatment-emergent subtypes of prostate cancer. For this work, three factors are critical: sample size, data depth, and platform flexibility. Tempus provides the scale needed to accurately identify and verify rare phenotypes. The multimodal data is also essential, as many cancer subtypes are defined by their transcriptome, not just genetics. Having the flexibility to explore the dataset with our own code in Workspaces allows us to refine our definitions and identify reliable signals in a way that is simply not possible with static datasets.
Wilson Ngai, MSc, PharmD, MBA: In medical affairs, my role is to generate data that fills evidence gaps in clinical practice. We recently completed a study comparing two treatments in patients with BRAF-mutated differentiated thyroid cancer, which is a very rare population. The core question—how our drug performs against another option in a molecularly defined, real-world population—is difficult to answer with traditional data sources like claims databases, which often lack progression or detailed molecular data. The Lens platform provided a rich clinical-genomic dataset that allowed us to identify the right patient cohort, identify an appropriate treatment option by line of therapy, and evaluate endpoints like real-world progression-free survival.
“This platform allows us to generate RWD addressing an important but unanswered question in clinical practice, in a very short period of time. We're talking about a couple of months, not years.”
– Wilson Ngai, MSc, PharmD, MBA, Senior Medical Director, Eisai
What were the major hurdles in conducting this type of research before integrated platforms were available?
Olivier Harismendy, PhD: A few years ago in academia, I tried to do this kind of research using our own institutional electronic health record (EHR) data. It was fragmented, siloed to a single hospital, and full of red tape. Aggregating data from other institutions was competitive and difficult. Furthermore, the molecular data we were generating, like Tempus reports, existed only as PDFs in the EHR and wasn't searchable. A platform like Lens solves this by aggregating and structuring data from across the country.
Wilson Ngai, MSc, PharmD, MBA: The traditional process was incredibly slow. To test a research idea, I would have to submit the hypothesis to various research partners, wait for them to conduct a feasibility analysis, and then evaluate the cost and timeline. The whole process could take weeks or months just to decide if a project was viable. Manual chart reviews are also not feasible for rare cancers, as it would take many months to find enough patients. An integrated platform lets us assess feasibility and analyze a large cohort very quickly—in one case, evaluating 10 to 20 research ideas in parallel and returning outcomes within a week.
How does the platform handle complex, unstructured data like clinical notes or imaging, and what's on the horizon for these modalities?
Shelley MacNeil, PhD: This is an area where AI is incredibly valuable. Tempus uses natural language processing to extract information from thousands of clinical notes at once, all within a secure, HIPAA-compliant environment. For imaging, Tempus has digital pathology images for the biopsies we sequence in-house, and we use AI to predict biomarkers or response directly from those images. Tempus is also beginning to integrate radiology images and apply AI to analyze them, which is an exciting area of future development. This allows researchers to access deeper insights from data that is traditionally very difficult to work with at scale.
As AI tools become more integrated into research, what are the key considerations for governance and successful adoption?
Olivier Harismendy, PhD: In biopharma, we are particularly careful and use AI only in a secure, enterprise context. We need to control who sees what data to maintain the integrity of blinded studies and protect internal information. In the context of a platform like Lens, I am excited about the potential for AI agents that can help design a study, from finding the data to performing the statistical analysis.
AI models can sometimes be a "black box." How does the Lens platform ensure the reproducibility and trustworthiness of its AI-generated insights?
Katie Igartua, PhD: This is a critical question, and our approach is very deliberate. We are creating and benchmarking AI agents for very specific, validated tasks. This allows us to understand their performance and limitations before we launch them. The tools you see today are deterministic, executing validated functions. As we release more advanced AI capabilities, our goal is to make the entire workflow transparent, allowing you to see the code and reproduce the full analysis from the user interface all the way to the workspace. Trust and reproducibility are paramount.
Note: Please note that the content in this document has been revised for clarity and conciseness. Some language and formatting may have been adjusted to enhance readability while preserving the original meaning and intent of the discussion.
To gain more insights about how Tempus Lens is advancing precision medicine, watch the full webinar recording or book a demo to see what the platform can do for your programs.
Article
