Skip to content

Evaluate Unconditional

Runs unconditional evaluation pipeline on the outputs of Alphafold or Colabfold models. Selects a folding that is most similar to the generation and returns a scores table.

Evaluate Unconditional

Quick Start

  1. Connect generated proteins (PDB) to generation_pdb input.
  2. Connect foldings (PDB) to folding_pdb input.
  3. Run the node to obtain the best folding and evaluation scores.

Setup Guide

1. Prepare Inputs

  1. Generate proteins using a diffusion or design node.
  2. Obtain foldings from Alphafold or Colabfold nodes.

2. Connect and Run

  1. Connect the generated proteins to generation_pdb.
  2. Connect the foldings to folding_pdb.
  3. Execute the node to receive evaluation results.

Basic Usage

Unconditional Evaluation

  • Compare generated proteins to predicted foldings.
  • Select the best matching folding for each design.
  • Output a CSV table with evaluation metrics (pst, rmsd, pae, plddt).

Configuration

Required Inputs

Field Description Type Example
generation_pdb Generated proteins to evaluate. PDB {"design_0": "...PDB..."}
folding_pdb Foldings to evaluate proteins against. PDB {"design_0_ranked_0": "...PDB..."}

Optional Inputs

None

Outputs

Field Description Example
best_folding.pdb Best of the predicted foldings. {"design_0": "...PDB..."}
score.csv Scores of the foldings (pst, rmsd, pae, plddt). "design_0,0.95,1.2,10.5,85.0"

Best Practices

Input Preparation

  • Ensure that generation_pdb and folding_pdb have matching or corresponding IDs for accurate evaluation.
  • Use outputs from supported generation and folding nodes for compatibility.

Efficient Evaluation

  • For large batches, consider running in TEST mode to limit evaluation to the first structure for faster feedback.

Troubleshooting

Common Issues

  • Mismatched IDs: Ensure that the IDs in generation_pdb and folding_pdb correspond to the same designs.
  • Empty Outputs: Check that both inputs contain valid PDB data.

Need Help?

  • Contact support for further assistance.