- Re-scaffold repo
- Better way of structuring both analysis and implementation of machine learning models (including already implemented agents).
- Data
- PRSIM secondary LFC input data download and wrangling
- Additional data downloading/wrangling on the fly (for augmenting existing data/acquiring pre-clinical/clinical labels)
- Implement machine learning models for adaptive screens
- Implement randomized tree regressor predictive of viability reduction from chemcial fingerprint
- Implement bayesian dose response model predictive of vaibility reduction and uncertainty from just screen outcomes
- Refactor agentic models to serve as additional base lines