An error occurred when running protein-small_molecule

[Jennyhu666]

Hello, thank you for publishing boltzgen!
I met problems when predicting protein-small fractions.
This is my yaml file:
entities:

• file:
  path: /home/biyun/Apps/boltzgen/test/prep.pdb.cif
  include:
  - chain:
  id: A
  - chain:
  id: B
  - chain:
  id: C
• ligand:
  id: D
  smiles: 'xxx'

An error occurred when I entered the following command:
/home/biyun/Apps/boltzgen/boltzgen_env/bin/boltzgen run ./xxx.yaml --output result --protocol protein-small_molecule --num_designs 10 --budget 2 --moldir /home/biyun/Apps/boltzgen/checkpoint/mol-file --folding_checkpoint /home/biyun/Apps/boltzgen/checkpoint/boltz2_conf_final.ckpt --affinity_checkpoint /home/biyun/Apps/boltzgen/checkpoint/boltz2_aff.ckpt --use_kernels false --design_checkpoints /home/biyun/Apps/boltzgen/checkpoint/boltzgen1_diverse.ckpt /home/biyun/Apps/boltzgen/checkpoint/boltzgen1_adherence.ckpt --inverse_fold_checkpoint /home/biyun/Apps/boltzgen/checkpoint/boltzgen1_ifold.ckpt

Traceback (most recent call last):
File "/home/biyun/Apps/boltzgen/boltzgen_env/bin/boltzgen", line 7, in
sys.exit(main())
^^^^^^
File "/home/biyun/Apps/boltzgen/src/boltzgen/cli/boltzgen.py", line 1700, in main
run_command(args)
File "/home/biyun/Apps/boltzgen/src/boltzgen/cli/boltzgen.py", line 566, in run_command
execute_command(args)
File "/home/biyun/Apps/boltzgen/src/boltzgen/cli/boltzgen.py", line 807, in execute_command
subprocess.check_call(command)
File "/home/biyun/Apps/boltzgen/boltzgen_env/lib/python3.11/subprocess.py", line 413, in check_call
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['/home/biyun/Apps/boltzgen/boltzgen_env/bin/python3.11', '/home/biyun/Apps/boltzgen/src/boltzgen/resources/main.py', 'result/config/design_folding.yaml']' returned non-zero exit status 1.

Looking forward to your reply.

[juliocesar-io]

Your CLI wrapper try/catch is not showning the full eror log but I assume is the same error I had for small molecule example, It's fixed on this PR: #167

Here is the linked issue: #166

[buerbaumer]

As @juliocesar-io mentioned, this is likely related to #166.

I tested protein-small_molecule protocol with PR #167 applied and it completes successfully.

Environment:

• AWS EC2 g5.2xlarge (NVIDIA A10G)
• BoltzGen via Docker with PR Fix analysis crash when SASA atom selection is empty - Small molecule example #167

Test:

boltzgen run example/small_molecule_from_file_and_smiles/4g37.yaml
--output test_171
--protocol protein-small_molecule
--num_designs 2
--devices 1
Result: All 7 pipeline steps completed:

✓ Step design completed successfully in 145.2s
✓ Step inverse_folding completed successfully in 11.1s
✓ Step folding completed successfully in 208.5s
✓ Step design_folding completed successfully in 208.6s
✓ Step affinity completed successfully in 212.0s
✓ Step analysis completed successfully in 17.5s
✓ Step filtering completed successfully in 10.5s´´´

@Jennyhu666 Your error traceback is truncated by the subprocess wrapper. The underlying error may be the same SASA computation issue fixed by PR #167. You can test by installing from the PR branch:

pip install git+https://github.com/HannesStark/boltzgen.git@refs/pull/167/head

[Jennyhu666]

Thank you for your reply. However, even after I installed this, there were still errors. Below are all the error messages I received. Could you please help me figure out where the problem lies?

error-allstep.log

[buerbaumer]

@Jennyhu666 I need to correct my earlier suggestion - this error is not related to PR #167 (the SASA fix). After analyzing your full error log, the actual issue is different.

Root Cause

Your YAML configuration doesn't specify any residues for design. The error:

File "data.py", line 838, in extract_residues
atom_indices = np.concatenate(...)
ValueError: need at least one array to concatenate

occurs because BoltzGen tries to extract designed residues at Step 4 (design_folding), but finds none.

Your Current YAML (missing designed residues):

entities:
  - file:
      path: /home/biyun/Apps/boltzgen/xxx/prep.pdb.cif
      include:
        - chain: {id: A}
        - chain: {id: B}
        - chain: {id: C}
  - ligand:
      id: D
      smiles: 'xxx'

This only includes existing chains and a ligand, but doesn't mark anything for design.

Solution

You need to either design a new binder protein OR redesign existing residues:

Option A - Design a new binder protein (recommended for protein-small_molecule):

entities:
  - protein:
      id: E                     # New chain for designed binder
      sequence: 140..180        # Design 140-180 residues
  - file:
      path: /home/biyun/Apps/boltzgen/xxx/prep.pdb.cif
      include:
        - chain: {id: A}
        - chain: {id: B}
        - chain: {id: C}
  - ligand:
      id: D
      smiles: 'xxx'

Option B - Redesign specific residues from the existing structure:

entities:
  - file:
      path: /home/biyun/Apps/boltzgen/xxx/prep.pdb.cif
      include:
        - chain: {id: A}
        - chain: {id: B}
        - chain: {id: C}
      design:                   # <-- ADD THIS BLOCK
        - chain:
            id: A
            res_index: 50..100  # Which residues to redesign
  - ligand:
      id: D
      smiles: 'xxx'

Verify Your Configuration

Run this to check if designed residues are detected:

boltzgen check ./xxx.yaml
Look for output like Total designed residues: N. If N is 0, your YAML needs a protein: entity or a design: block.

Reference

See working examples:
example/protein_binding_small_molecule/brilacidin.yaml - simple protein + ligand
example/small_molecule_from_file_and_smiles/4g37.yaml - with existing structure + ligand

Sorry for the confusion with my earlier comment about PR #167 - that fix addresses a different issue (SASA computation crash during analysis, not missing designed residues).

[Jennyhu666]

Thank you very much for your reply. It has been very helpful to me. I will revise the YAML file again.