New Barcelona Spin-off HelixAI Leverages Graph Foundation Models to Convert Complex Biomedical Data Into Clinical Insights

IRB Barcelona and UPC launch HelixAI, using Graph Foundation Models to analyze complex multi-omic data for oncology and personalized longevity insights.

By: AXL Media

Published: Apr 30, 2026, 8:09 AM EDT

Source: Information for this report was sourced from IRB Barcelona

Bridging the Knowledge Gap Between Biological Data and Actionable Medicine

While the field of biomedicine currently generates an unprecedented volume of raw data, the ability to transform these complex sets into practical clinical knowledge remains a significant bottleneck. To solve this disparity, IRB Barcelona, ICREA, the Universitat Politècnica de Catalunya (UPC), and the tech firm Napptilus have established HelixAI. This new spin-off is dedicated to building artificial intelligence infrastructure specifically designed to navigate the intricate landscape of biomedical analysis. By integrating multi-omic datasets, the founders intend to create a clearer path for researchers and clinicians to move from laboratory observations to bedside applications.

Assembling a Multi-Disciplinary Leadership Team for Technological Growth

The foundation of HelixAI rests on a strategic blend of scientific expertise and business development experience. The company was co-founded by Dr. Salvador Aznar Benitah, a prominent ICREA researcher at IRB Barcelona, alongside Prof. Pere Barlet from UPC and Rafa Terradas of the Napptilus tech lab. This leadership structure ensures that the firm’s technological outputs are grounded in rigorous scientific inquiry while remaining viable for the commercial market. The collaboration represents a major effort to place advanced computational tools at the center of the Catalan biomedical ecosystem, drawing from academic and private sector strengths.

Deploying Graph Foundation Models for Heterogeneous Biological Datasets

Traditional artificial intelligence models often struggle with the unique constraints of biomedical data, which is frequently high-dimensional yet limited in sample size. To address these limitations, HelixAI is developing proprietary technology based on Graph Foundation Models. According to Prof. Pere Barlet, this generation of AI is uniquely capable of detecting hidden relationships across transcriptomics, metabolomics, and lipidomics. Unlike standard AI applications that require massive, uniform datasets, these specialized models are engineered to extract reliable conclusions from the diverse and often small-scale samples typical of real-world clinical environments.