Drug discovery using bioactivity data from ChEMBL. This has been adapted from the repository "CDD_ML_Part_*_bioactivity_data.ipynb" from @dataprofessor. I made some small improvements:
- Reduced noise via removal of flagged samples
- Data augmentation via addition of samples reported in Ki (rather than just IC50 by applying the conversion factor reported in Kalliokoski et al. (DOI.0061007).
- More plots in data analysis exploration
- Kruskal-Wallis and Conover-Iman tests
Future plans for the project include structural analysis using Bio3D and the SMILE IDs reported in part 1 of the code, as well as a docking simulation between the top 3 samples and the target (Abeta).
I also intend to condense the various parts of the code into one or two notebooks instead of five. I will try and balance brevity and clarity as well.