How Ardigen’s daGama framework helped a drug discovery company identify crucial genetic targets for cancer treatment?
Identifying essential genes in cancer is a cornerstone of targeted drug discovery. Yet, the integration of diverse multi-omics datasets and the complexity of advanced machine learning methods pose significant challenges. In this case study, we explore how a growing drug discovery company partnered with Ardigen to explore the use of AI and multi-omics data integration for breakthrough insights into gene essentiality.
Challenge
The client faced three key hurdles:
- Lack of in-house Machine Learning expertise: Despite having rich omics datasets from patients and cell lines, the client lacked the ML capabilities to extract actionable insights.
- Data heterogeneity: Combining in vivo and in vitro multi-omics profiles introduced batch effects and inconsistencies.
- Need for rapid, reliable outcomes: Time-sensitive research required fast, yet scientifically robust, results.
Approach
To tackle these challenges, Ardigen deployed its proprietary daGama Framework, blending computational biology expertise with cutting-edge machine learning methods. Key elements of the approach included:
- Variational Autoencoders (VAE): Used for learning compact, informative representations of genetic profiles.
- Multimodal VAE (MVAE): Integrated various omics data types into a unified, consistent format.
- Batch effect correction and domain adaptation: Ensured data harmonization across sources, improving cross-domain reliability.
This robust strategy enabled the creation of a harmonized, multi-dimensional view of gene activity across both patient-derived and cell line datasets.
Results
- >70% improvement in predictive performance with domain adaptation, highlighting the importance of advanced correction techniques.
- 10+ synthetically lethal gene pairs identified, including novel discoveries and known pairs, validating the method’s accuracy.
- Successfully identified essential genes across multiple cancer types, delivering a critical foundation for targeted therapy research.
Through collaboration with Ardigen and the application of its daGama Framework, the client transformed a fragmented data challenge into a high-impact research success. This project not only enhanced the state of the art in gene essentiality prediction but also accelerated the client’s drug discovery efforts with scientifically validated insights.
Expert Contribution
Reviewed by: Dr. Marek Kudła, PhD
Role: Senior Scientist, AI-Driven Drug Discovery
Expertise: Small molecule drug design, computational chemistry, machine learning modeling
