Introducing GENAVA

🧬 Introducing GENAVA — WeDaita's biology-native AI foundation model for virtual target validation.

Drug discovery starts with a costly question: What happens when you perturb a gene? Answering that question traditionally requires months of experiments and significant capital.

Built on a biological pathway knowledge graph, GENAVA predicts not only whether a target matters, but also which pathways and processes are affected, providing interpretable biological reasoning instead of a black-box score.

We validated GENAVA against the Broad Institute DepMap CRISPR knockout dataset (1,208 cell lines × 18,531 genes). Despite being trained on 75 human tissue samples, its predictions still showed a significant correlation with experimentally measured fitness effects (p = 0.006) on unseen data from cell lines.

We also confirmed a known drug mechanism: TNF inhibition correctly disrupted inflammatory pathways relevant to inflammatory bowel disease.

Early results, but a promising step toward faster, more explainable target validation in drug discovery. More tissue types and disease-specific fine-tuning of our AI model is on the roadmap. Keep tuned!

--------------
♻️ Know someone who may benefit from accelerating drug discovery with AI?
➡️ Feel free to share: https://lnkd.in/esuPivNr
---------------
#DrugDiscovery #AI #Bioinformatics #ComputationalBiology #Biotech #FoundationModels #WeDaita #BioMedAI #KnowledgeGraph #BiologicalPathway