Spring 2021 - Automation of Scientific Research - course project
- Built an active learning regression model to predict binding affinity between MHC class I and small peptides and compared the results against an offline learner (Random Forest).
- Tune model hyperparameters using grid search.
- Using Bayesian optimization, simulated a peptide design experiment trying to find peptides with high binding affinity to MHC class I within a stringent query budget.
- Created a Bayesian optimizer with Gaussian process as the regressor and max expected improvement as the queuing strategy and compared it with a Bayesian optimizer with random forest as regressor.
- Utilized Jupyter Notebook and Python (scikit-learn, numpy, matplotlib, modAL).