AI supermodel set to predict cancer drug response

Embracing the promise of AI as an aid in cancer research, Slovenian-US biotech firm Genialis is addressing a key challenge in cancer treatment, the lack of reliable biological markers to predict how patients will respond to therapy, with an advanced AI model, developed from analysing over a million global RNA sequencing samples.

The company was founded nine years ago as a spin-out of the Bioinformatics Laboratory of the Faculty of Computer and Information Science (FRI), University of Ljubljana. Genialis is headquartered in Boston but retains strong ties with Slovenia, employing 22 Slovenian experts in computing, maths, physics, medicine, and biology, CE Report quotes The Slovenia Times.

The company develops software tools that are used in new cancer drug development, treatment planning, and diagnostics. They develop biomarkers that capture complex cancer biology from gene expression data, transforming raw RNA sequencing data into actionable insights with predictive AI.

Miha Štajdohar, Genialis co-founder and CTO, says Slovenian expertise was vital to the company's creation. "Slovenia has a very strong ecosystem in AI. We have renowned laboratories at the FRI and the Jožef Stefan Institute, where groups have been developing AI knowledge for decades, and this tradition also enables the development of quality personnel," he pointed out at a recent event in Ljubljana.

Initially developing software for biologists, Genialis pivoted six years ago towards personalised medicine. "We upgraded our software with AI tools that offer diagnostic and pharmaceutical companies answers to specific questions, for example, whether a patient will respond to a specific cancer drug or therapy," Štajdohar explained.

The absence of reliable biomarkers hinders effective care. Genialis tackles this with its proprietary "Genialis Supermodel", an AI model using extensive global patient RNA data, co-founder and CEO Rafael Rosengarten said. Machine learning models then predict drug efficacy for individual patients.

Key challenges however persist. Rosengarten cited accessing diverse, quality patient data as crucial, leading to partnerships across India, the Middle East, Europe, and Taiwan to build a more inclusive database.

Funding is another significant hurdle, as research is costly, with 96% of oncology drugs never reaching the final phase of clinical testing. Genialis is currently seeking its next investment round.

Štajdohar is confident the future lies in targeted therapies enabling effective, individualised plans. He explained that their biomarkers significantly boost success: while previously only 20-30% might respond, patient selection using their biomarkers means treatment success rises to 65%.

Currently used in clinical studies and research clinics, the ultimate aim is for their software and diagnostic tests to become part of routine clinical practice, directly aiding patients and improving their quality of life.