Evaluate Unconditional¶

Evaluates a set of generated protein structures against one or more folded structures and selects the best folding per design. It also produces a scoring table with key structural metrics (pst, rmsd, pae, plddt) for review. Supports MOCK, PROD, and TEST modes to balance speed, realism, and cost.

Usage¶

Use this node after generating candidate protein structures and obtaining corresponding foldings. Provide both the generated structures and the foldings; the node will compare them, pick the best folding for each design, and output a CSV with evaluation scores. Ideal in workflows where you need an automated selection of best foldings and a quantitative summary of structure quality.

Inputs¶

Field	Required	Type	Description	Example
generation_pdb	True	PDB	Generated protein structure(s) to evaluate. Accepts one or multiple entries, typically a mapping from design name to PDB content.	{"design_A": "PDB_CONTENT_STRING_OR_PATH", "design_B": "PDB_CONTENT_STRING_OR_PATH"}
folding_pdb	True	PDB	Predicted folding structure(s) to evaluate the generated proteins against. Provide one or more PDBs, typically a mapping from folding name to PDB content.	{"design_A_ranked_0": "PDB_CONTENT_STRING_OR_PATH", "design_B_ranked_0": "PDB_CONTENT_STRING_OR_PATH"}
mode	True	CHOICE	Execution mode. Options: MOCK (returns predefined sample results for quick demos), PROD (runs the full backend evaluation), TEST (runs a minimal version, limiting evaluation to the first structure in each input for faster turnaround).	PROD

Outputs¶

Field	Type	Description	Example
best_folding.pdb	PDB	The best folding selected per input design. Returned as a mapping from design name to PDB content.	{"design_A": "PDB_CONTENT_STRING", "design_B": "PDB_CONTENT_STRING"}
score.csv	CSV	A CSV string summarizing evaluation metrics per structure (includes columns such as pst, rmsd, pae, plddt).	name,pst,rmsd,pae,plddt\ndesign_A,0.85,1.9,2.3,92.1\ndesign_B,0.79,2.4,3.1,89.5

Important Notes¶

Mode behavior: MOCK returns canned data; PROD runs the full evaluation; TEST limits processing to the first structure in each input to reduce runtime.
Provide at least one PDB in both generation_pdb and folding_pdb. If many structures are provided, processing time scales accordingly.
Output best_folding.pdb contains one entry per evaluated design, mapping the design name to its selected best folding.
The score.csv output is a CSV string; save it to a file or parse it downstream if you need structured access to metrics.
Ensure inputs are valid PDB contents or resolvable PDB payloads as accepted by your environment.

Troubleshooting¶

No results or empty outputs: Verify both generation_pdb and folding_pdb contain at least one valid PDB entry each.
Unexpectedly long runtime: Use TEST mode to validate the setup or reduce the number of input structures.
Results look synthetic: Ensure mode is set to PROD instead of MOCK.
Missing best folding for a design: Confirm that the folding_pdb input includes the relevant foldings for the corresponding generated design names.
CSV parsing issues: Remember score.csv is returned as a string; write it to a .csv file or parse it with a CSV reader.