Domain-adaptive foundation model bridges clinical and preclinical cancer biology across 130,000+ oncology samples

CAMBRIDGE, MA / ACCESS Newswire / April 21, 2026 / Data4Cure, Inc., the company behind the Biomedical Intelligence® Cloud for pharmaceutical R&D, today announced the upcoming presentation of two late-breaking poster abstracts at the 2026 American Association for Cancer Research (AACR) Annual Meeting in San Diego, CA. The posters describe RNA1-DA, a new domain-adaptive extension of the company's RNA1 foundation model, designed to enable forward and reverse translation between clinical tumors and preclinical cancer models.

RNA1-DA: Bridging Clinical and Preclinical Cancer Biology

Preclinical cancer models - cell lines, organoids, and patient-derived xenografts (PDX) - are essential tools for drug discovery, but they differ from patient tumors in cellular composition, molecular profiles, and environmental context. These differences make it challenging to translate findings between systems.

RNA1-DA addresses this challenge by extending Data4Cure's RNA1 foundation model with a deconvolution layer for separating cancer cell expression from bulk tumor samples and a domain adaptation layer that aligns sample embeddings across clinical and preclinical systems.

In the first poster (Abstract LB434), Data4Cure researchers integrated over 130,000 clinical and preclinical samples and demonstrated that RNA1-DA enables molecular subtype transfer between clinical and preclinical systems, data-driven preclinical model selection, and improved drug response prediction.

Collaboration with Champions Oncology

In the second poster (Abstract LB435), Data4Cure applied RNA1-DA to a cohort of over 1,500 low-passage PDX models from Champions Oncology's TumorGraft bank. By embedding these samples jointly with 130,313 public clinical and preclinical samples from the Data4Cure Oncology Sample Universe, the study demonstrated accurate disease classification, concordant molecular subtype assignment, and recapitulation of 87% of subtype-mutation associations from TCGA in the Champions PDX cohort. Multiple PDX-transferred subtypes also showed significant associations with in vivo drug responses that align with known clinical patterns.

"A key challenge in translational oncology is that preclinical models and patient tumors live in different molecular spaces. RNA1-DA learns to bridge that gap, so that molecular subtypes, genetic dependencies, and drug response signals can be transferred across systems in a principled way," said Janusz Dutkowski, Ph.D., Founder and CEO of Data4Cure. "The collaboration with Champions Oncology validates this approach on a large, independent PDX cohort. Our results support the research use of RNA1-DA in studies focused on forward/reverse translation, patient stratification, and biomarker discovery."

Poster Presentation Details

Both posters will be presented on April 22, 2026, 9:00 AM - 12:00 PM, Section 52, at the AACR 2026 Annual Meeting in San Diego, CA.

• Abstract LB434: RNA1-DA: A domain-adaptive RNA foundation model for forward and reverse translation. Edward O'Brien, Michał Kukiełka, Aleksandra Cupriak, Roy Ronen, Janusz Dutkowski. Data4Cure, Inc.

• Abstract LB435: Large-scale foundation model-based PDX model selection and cancer subtype assignment. Edward O'Brien, Alex Moreau, Gervaise Henry, Gilad Silberberg, Tammer Farid, Roy Ronen, Janusz Dutkowski. Data4Cure, Inc. and Champions Oncology.

About Data4Cure

Data4Cure, Inc. is the company behind the Biomedical Intelligence® Cloud - an integrated platform combining a knowledge graph with over 4 billion provenance-tracked biological relationships (CURIE), hundreds of thousands of harmonized biological samples, biology-focused foundation models, and AI-driven analytics to transform biomedical data into actionable knowledge for drug discovery. The platform is used by leading pharmaceutical companies for target and biomarker discovery, patient stratification, and translational research. Data4Cure is headquartered in Cambridge, MA. For more information, visit data4cure.com.