Dani Bodor

Make `MapMethod` an attribute of `GridSettings` in deeprank2.grid module

1

This will simplify code in many places throughout. Also, consider making augmentation count a `GridSettings` attribute, but this is less logical, as it is not truly a setting of the...

make `_set_feature_modules` to return `ModuleType` objects

1

see https://github.com/DeepRank/deeprank2/pull/492#discussion_r1337177705 in query.py # TODO: I would prefer if `_set_feature_modules` would return a list of `ModuleType` instead of `str`, # but this leads to an exception in the pool...

Create feature classes

2

Build a parent class for DeepRankFeature and then use this to create child classes for each independent feature, that contain an `add_feature` function. This allows for much better control when...

variant residue is only added to atomic graphs, but not residue graphs

1

See: https://github.com/DeepRank/deeprank2/pull/492#issuecomment-1732423764

evaluate whether we want to stop suppressing errors by default in `_process_one_query` method of `QueryCollection`

3

In principle, we expect users to have "good" PDB files. Currently, we are catching errors in case there is a problem with the pdbs. Evaluate whether if all PDBs are...

Variant queries requesting redudant inputs

5

Currently, we are requesting as inputs for variant queries (among other things): number, insertion code, wt aa, var aa. At least one of those is redundant because it can be...

Use openMM to redefine the forcefield

5

This has multiple advantages: - allows optional (and correct) generation of hydrogens in pdb structures based on predicted protonation state - allows determining 1-3 and 1-4 pairings for nonbond energy...

Reorganize `Trainer` class

6

Although everything works as intended now, and is reasonable logical, it would be nice to adhere to the logic of pytorch lightning (see: [here](https://github.com/Lightning-AI/lightning/blob/master/src/pytorch_lightning/trainer/trainer.py)) Reorganize which arguments are passed to...

Check performance warnings of CNNs

3

test_trainer gives the following warnings. We should investigate and either disable them or make the recommended fixes. > tests/test_trainer.py::TestTrainer::test_grid_regression > /home/dbodor/git/DeepRank/deeprank-core/deeprankcore/dataset.py:376: UserWarning: Creating a tensor from a list of numpy.ndarrays...