x
[contact-form-7 404 "Not Found"]

Nov2025

Ariana Pharmaceuticals Publishes New Study on AI-Driven Discovery of Genomic Biomarkers for Liver Toxicity Risk

Ariana Pharmaceuticals is proud to announce the publication of a new Real-World Evidence study: “eXp AI-driven Discovery of Genomic Biomarkers for Liver Toxicity Risk - A UK Biobank Case Study.” in The Pharmacogenomics Journal

This research showcases how Ariana’s explainable Artificial Intelligence (eXp AI) and proprietary KEM® (Knowledge Extraction & Management) platform can identify genomic biomarkers associated with mortality risk in toxic liver disease from the analysis of data from the UK Biobank RWE data base , paving the way for safer and more personalized medicine. 

Re-Imagining Safety Biomarkers with Explainable AI

    • In drug development, identifying safety biomarkers-those that predict adverse reactions such as liver toxicity-has long been a challenge. Unlike efficacy biomarkers, which are more widely used, validated safety biomarkers remain scarce.  Ariana’s new study redefines what’s possible. By leveraging eXp AI, a logic-based approach that goes beyond traditional statistical models, researchers uncovered meaningful and interpretable genetic insights that could guide risk stratification, patient selection, and drug safety evaluation.  “Our study demonstrates how explainable AI can uncover rare but critical genetic variants that influence survival in liver toxicity,” said Federico Goodsaid, Ariana‘s Regulatory Affairs SVP. “Traditionally, liver safety assessments in clinical trials have been drug-centered. This research shifts the paradigm toward patient-centric safety biomarkers-helping identify individuals vulnerable to severe adverse liver events before exposure. By applying logic-based algorithms, we transform complex genomic data into clinically actionable insights that advance both safety and precision in drug development.”

About the study

  • The study analyzed genomic and clinical data from 225 patients diagnosed with toxic liver disease within the UK Biobank cohort of over 500,000 participants. Using KEM®, Ariana’s AI integrated survival outcomes, phenotypes, comorbidities, and 36,394 liver-related SNPs (single nucleotide polymorphisms) derived from key biological pathways including: 
        • ● KEGG Non-Alcoholic Fatty Liver Disease (NAFLD) pathway.
        • ● Adverse Outcome Pathway (AOP) models for liver injury.
    Through advanced AI-driven association and concept analysis, Ariana identified 15 genomic variants strongly linked to increased mortality in toxic liver disease. Among them, one SNP- rs73158145 in the PRKAG2 gene—was found to reduce median survival by 3.5 times in homozygous carriers.

Key-findings

      • ● 15 significant SNPs associated with mortality in toxic liver disease
      • ● 85% mean accuracy achieved by the KEM®-selected model, outperforming traditional methods (68.9%).
      • ● Sensitivity of 66.2% and specificity of 91.8%, demonstrating high predictive reliability.
      • ● Variants showed hazard ratios between 2.55 and 8.91, highlighting strong prognostic value.
These findings illustrate how AI-enhanced biomarker discovery can illuminate genomic signatures that were previously undetectable through conventional analytics.

Impact on drug safety and development

          • This work marks a significant milestone in pharmacogenomic safety biomarker discovery. By combining real-world evidence (RWE) with explainable AI, Ariana demonstrates a pathway toward integrating genomic safety biomarkers into clinical and regulatory decision-making.
          Such integration could dramatically reduce adverse outcomes and drug withdrawals due to hepatotoxicity, improving both patient safety and R&D efficiency.

Ariana’s vision: Depth over volume

      • At Ariana, we believe the depth of your data matters more than its volume. Our AI does not rely on large datasets—it thrives on rich, multi-dimensional data, capturing both strong and subtle patterns that reveal the true drivers of clinical outcomes. 
        • ● Identify safety and efficacy biomarkers early in development.
        • ● Optimize dosing and patient selection
        • ● Integrate real-world data to simulate control arms and enhance clinical relevance.
        • ● Build interpretable models that support regulatory engagement and precision medicine

Towards a Safer, More Personalized Future

            • This study represents not just scientific advancement, but a reaffirmation of Ariana’s mission – to make drug development more intelligent, explainable, and patient-centered.
        By combining real-world-evidence, genomics, and explainable AI, Ariana continues to bridge the gap between discovery and clinical application, unlocking insights that help bring the right therapies to the right patients- safely and effectively.
          Such integration could dramatically reduce adverse outcomes and drug withdrawals due to hepatotoxicity, improving both patient safety and R&D efficiency.
Link to the publication