Identifying complex biomarkers at scale
Unlearn Omics uses unsupervised deep learning to discover complex biomarkers at previously impossible speed and scale.
First, we use a single model—trained on unlabeled data aggregated from all potential sources—that continuously learns to disentangle the relevant sources of variation in the data (which we call “features”).
Then, we use small targeted studies (~100 subjects) with measured outcomes to determine which features predict the outcome.
Learning features to identify phenotypes
Benefits of Unlearn Omics
- Leverages shared biology across disease areas
- Generalizes better between cohorts (less overfitting)
- Continuously improves as more data is collected