deepchem: Errors featurizing PDBBind with RdkitGridFeaturizer

I’m trying to run this script from the book examples, which uses RdkitGridFeaturizer to featurize PDBBind. Here’s what I find.

• With the latest code, using the core subset and featurizing only the binding pockets, it is much faster than before (minutes instead of hours).

• While running, it fills the console with a huge number of lines like these:

Coordinates are outside of the box (atom id = 7, coords xyz = [-3.97537037  0.55992593 -8.72777778], coords in box = [ 2  4 -1]
Coordinates are outside of the box (atom id = 105, coords xyz = [-1.53037037 -1.94707407 -8.81577778], coords in box = [ 3  3 -1]
Coordinates are outside of the box (atom id = 106, coords xyz = [-1.41437037 -0.88707407 -9.36477778], coords in box = [ 3  3 -1]

I suspect there’s one line for every atom that isn’t in the binding pocket?

• It fails with this error:

/home/peastman/miniconda3/envs/tf2/lib/python3.8/site-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  return array(a, dtype, copy=False, order=order)
Traceback (most recent call last):
  File "pdbbind_nn.py", line 11, in <module>
    n_features = train_dataset.X.shape[1]
IndexError: tuple index out of range