[BUG] RFD3 might have problems in reading PDB format file with HETATM inside.

[Huilin-Li]

Describe the bug
My target protein has a phospho group at position 12. I use its PDB format file as the input. But I got an error like:

ValidationError: 1 validation error for DesignInputSpecification
  Value error, [component=A12] Residue A12 not found in atom array. [type=value_error, input_value={'input': '/home/huilinli...lse, atoms=0, tokens=0)}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.13/v/value_error

To Reproduce
Here, I use one of your example CD3e.cif to reproduce this bug:

1. open CD3e.cif in PyMol, and save a new file as CD3e.pdb
2. .yaml file is

hetatm_test:
  input: (path)/CD3e.pdb
  contig: 100-110,/0,A1-13
  redesign_motif_sidechains: false
  is_non_loopy: true

3. .sh file is

#!/bin/bash

# activate environment
source ~/miniconda3/etc/profile.d/conda.sh
conda activate RFD3_foundry

export LD_LIBRARY_PATH=$CONDA_PREFIX/lib:$LD_LIBRARY_PATH


rfd3 design \
  out_dir="(path)/test_outputs" \
  inputs="(path)/test.yaml" \
  n_batches=3 \
  diffusion_batch_size=1 \
  dump_trajectories=1

I got the error:

ValidationError: 1 validation error for DesignInputSpecification
  Value error, [component=A3] Residue A3 not found in atom array. [type=value_error, input_value={'input': '/home/huilinli...lse, atoms=0, tokens=0)}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.13/v/value_error

However, if I use CD3e.cif, RFD3 works.
Expected behavior
A clear and concise description of what you expected to happen.

Screenshots

CD3e.pdb

Additional context
Add any other details that are useful for understanding the problem here.

[rclune]

Sorry for the delayed response.

I took a look at your attached PDB file and it looks like the third residue is non-canonical, this is likely why you got this error message. The reason that this is occurring only when using the PDB input and not the CIF input is because they are being parsed slightly differently. The PDB parser treats the non-canonical amino acid as if it is in a separate chain (see _parse_from_pdb). CIF files are not parsed in the same way, so this issue is avoided.

Having said that, you can use chains with non-canonical amino acids as an input to RFD3, but it is highly encouraged that you keep the atoms in the non-canonical amino acid fixed.