feat: RandomLigandTransform to perform system level ligand transforma…

…tion (to be used in training) (#17)

* added  ()

* added structure transform

* separated target and feature transforms

* added type annotations

* docs: update readmes with latest PinderDataset/Loader APIs reflecting separate transforms applied to feature and target complex

* docs: update loader notebook w/latest transform API

* docs: (WIP) update MLSB nb with latest transforms API

* docs: update example notebooks w/latest output containing chain_ids in the structure2tensor transform

* bumpversion minor

---------

Co-authored-by: danielkovtun <[email protected]>

README.md

@@ -52,7 +52,7 @@ for `conda` for the same functionality):
 
 
 ```bash
-mamba create --name pinder python=3.10
+mamba create --name pinder python=3.11
 mamba activate pinder
 ```
 
@@ -451,7 +451,8 @@ train_dataset = PinderDataset(
  base_filters: list[filters.PinderFilterBase] = [],
  sub_filters: list[filters.PinderFilterSubBase] = [],
  structure_filters: list[filters.StructureFilter] = [],
- structure_transforms: list[transforms.StructureTransform] = [],
+ structure_transforms_target: list[transforms.StructureTransform] = [],
+ structure_transforms_feature: list[transforms.StructureTransform] = [],
 )
 
 train_loader = get_torch_loader(train_dataset, batch_size=2, shuffle=True)
@@ -834,7 +835,7 @@ git clone [email protected]:pinder-org/pinder.git
 
 ```bash
 cd pinder
-mamba create --name pinder python=3.10
+mamba create --name pinder python=3.11
 mamba activate pinder
 ```
 

examples/README.md

@@ -321,7 +321,7 @@ sub_filters = [
  filters.FilterSubRmsds(rmsd_cutoff=7.5),
  filters.FilterDetachedSub(radius=12, max_components=2),
 ]
-# We can include Structure-level transforms (and filters) which will operate on the target and feature complexes
+# We can include Structure-level transforms (and filters) which will operate on the target and/or feature complexes
 structure_transforms = [
  transforms.SelectAtomTypes(atom_types=["CA", "N", "C", "O"])
 ]
@@ -331,7 +331,10 @@ train_dataset = PinderDataset(
  monomer_priority="random_mixed",
  base_filters = base_filters,
  sub_filters = sub_filters,
- structure_transforms=structure_transforms,
+ # Apply to the target (ground-truth) complex
+ structure_transforms_target=structure_transforms,
+ # Apply to the feature complex
+ structure_transforms_feature=structure_transforms,
 )
 batch_size = 2
 train_dataloader = get_torch_loader(