| 1 |
compute and extract
all
geometric (clustering) and chemical features of putative pockets with
appropriate normalization |
| 2 |
load 2 pairs (geometry and
chemistry) of pretrained Isolation Forests (IFs): one trained on the
“large” pockets, and one trained on the “small”
(sub)pockets; the former is the main score, while the latter is used
only to compare subpockets with each other |
| 3 |
compute the anomaly score
from the (main) geometric and chemical IFs; rank all putative pockets
according to the average score |
| 4 |
if any, rank the subpockets
within each pocket |
| 5 |
return to the user the pockets
ranked according to point 3, and provide the subrank of subpockets,
if any, according to point 4 |
