On April 23, 2025 Genialis, the RNA biomarker company, reported two new studies demonstrating how predictive algorithms can drive better outcomes for patients, providers, and drug developers (Press release, Genialis, APR 23, 2025, View Source [SID1234652080]). In poster presentations at the American Association for Cancer Research (AACR) (Free AACR Whitepaper) Annual Meeting 2025, Genialis showed its AI foundation model of cancer—the Genialis Supermodel—can predict how a patient will respond to a specific therapy (Abstract 3313 / 16). The Supermodel can also help explain how drugs work in the body and how the tumor learns to resist treatment, as well as propose potential new drug combinations to improve outcomes. The AACR (Free AACR Whitepaper) presentations share these insights from Genialis’ own work with KRAS G12C NSCLC patient data and its collaboration with Debiopharm on the Debio 0123 WEE1 inhibitor in various cancer types. (Abstract 3659 / 21)
Schedule your 30 min Free 1stOncology Demo!
Discover why more than 1,500 members use 1stOncology™ to excel in:
Early/Late Stage Pipeline Development - Target Scouting - Clinical Biomarkers - Indication Selection & Expansion - BD&L Contacts - Conference Reports - Combinatorial Drug Settings - Companion Diagnostics - Drug Repositioning - First-in-class Analysis - Competitive Analysis - Deals & Licensing
Schedule Your 30 min Free Demo!
Most drug candidates fail in clinical trials, and of the ones that are successful, most cancer patients do not benefit, a problem that continues to plague drug developers, physicians, patients, and their families. Predictive biomarkers are known to improve these results considerably. The Genialis Supermodel is a large molecular model trained on over one billion RNA-seq-derived data points that enables rapid development and validation of accurate and information-rich biomarker algorithms across a wide range of cancer drug targets to act as a recommendation engine for every cancer target, drug, and patient.
"A shortcoming in precision medicine is we need a deeper understanding of individual patient biology and the molecular context in which a drug is acting, as this can provide crucial translational and clinical intelligence," said Rafael Rosengarten, PhD, CEO of Genialis. "We built the Genialis Supermodel to embrace the complexity and learn the underlying biology of cancer so we can guide therapy decisions and derisk drug development. At AACR (Free AACR Whitepaper), we’re showing that the same framework applies across very different drug targets and tumor types – and most importantly, that it works."
Genialis Supermodel for KRAS-mutated cancers
KRAS mutations drive approximately 25 percent of all human cancers, leading to an estimated 3.9 million new cases each year. Despite the recent approvals of two KRAS inhibitors (KRASi), these drugs show limited efficacy and durability. Response rates hover around 40 percent, and the median patient benefit lasted just 8 to 10 months, in part because the available biomarkers for patient selection rely on genotype alone. The Genialis Supermodel addresses this challenge by mapping over twenty biological modules relevant to KRASi-therapy in every tumor sample rather than focusing solely on the allele.
The first poster, led by Genialis, incorporates analysis of therapeutic mechanisms related to EGFR, immune checkpoint inhibitors (ICIs), and standard-of-care chemotherapy to stratify patients with KRAS-mutated cancers using the company’s first-in-class RNA biomarker algorithm, Genialis krasID. By broadening the biological space considered by the algorithm, Genialis is better able to predict results from both well-established and investigational therapies, to model various combinations, and to suggest the most effective sequencing/indication of various compounds. This new research expands on Genialis’ novel work presented at AACR (Free AACR Whitepaper) 2024. Genialis krasID serves as a sterling example of how the Genialis Supermodel scales to align treatment selection with the underlying tumor biology and therapeutic context.
"Genotype determines eligibility, but it often fails to predict actual clinical benefit. The Supermodel addresses this gap by capturing the functional biology that drives response," said Josh Wheeler, MD PhD, Director of Research and Innovation at Genialis. "The biology-first approach engendered by the Genialis Supermodel provides the mechanistic understanding and the necessary precision for informing and stratifying both emerging and standard of care therapeutic selection in patients."
Predicting response to WEE1 inhibitor Debio 0123
The second poster, presented in collaboration with Swiss-based global biopharmaceutical company Debiopharm, describes early results from the development of a predictive model for Debio 0123, a WEE1 inhibitor currently in phase 1 clinical trials. While the collaboration has since expanded to clinical studies, this poster describes the use of the Supermodel to identify biological modules in the DNA damage response space that contribute to Debio 0123 WEE1 response, training of a first-generation machine learning algorithm, and exploration of that biomarker across various translational models.
Using the DNA damage response (DDR)-related modules as input features, the ML algorithm was trained on patient-derived organoid datasets and tested on xenografts. The resulting biomarker algorithm achieved promising performance in leave-one-out cross-validation. Notably, it accurately predicted an in vivo responder that had been missed by organoid IC50 testing, showing the model’s ability to capture biologically relevant signals that may not be observed through traditional in vitro assays. These early findings set the foundation for further validation in independent clinical datasets and point to broader use of this biomarker framework for other DDR-targeting agents.
"The findings from this study show the Genialis Supermodel is able to learn components of complex biology that are translationally relevant and likely to prove important in human datasets. We see strong potential in using this kind of RNA-based, ML-driven approach to guide patient selection and boost development success," said Luke Piggott, Director, Global Business Development & Principal Scientist at Debiopharm. "The ability to predict, and soon validate, response is now being born out in clinical data and gives us an important strategic advantage."
These two AACR (Free AACR Whitepaper) posters demonstrate how Genialis supports drug development through its modular, interpretable Supermodel. Whether targeting KRAS-driven tumors or DDR mechanisms or entirely novel target biology, Genialis’ predictive biomarker algorithms reflect tumor biology, making it possible to match patients to successful treatment options and improve clinical outcomes.