| name | molclaw-evobind-tool |
| description | Design linear or cyclic peptide binders from receptor FASTA sequences using EvoBind2 with structured result outputs. |
| license | MIT license |
| metadata | {"skill-author":"PJLab"} |
EvoBind2 Peptide Binder Design
Note:
- Local files are not directly accessible by the server. Please upload them to the server using
drugsda-file-transfer before execution.
- For PDB file inputs, it is recommended to preprocess them using
drugsda-fix_pdb before execution.
Usage
2. EvoBind2 Binder Design
The description of tool evobind_tool.
Design linear or cyclic peptide binders from a receptor sequence using EvoBind2 in structure-guided screening workflows.
Args:
fasta (str): Receptor FASTA file path.
peptide_length (int): Binder peptide length, default 10.
num_designs (int): Number of independent design rounds, default 10.
num_iterations (int): Monte Carlo iterations per round, default 100.
max_recycles (int): AlphaFold2 recycle count, default 1.
model_name (str): AlphaFold2 model in {model_1, model_2, model_3, model_4, model_5}, default model_1.
target_residues (str): Receptor target residues as comma-separated 1-indexed positions or all, default all.
cyclic (bool): Whether to enable cyclic peptide design, default False.
msa_file (str|None): Optional precomputed MSA .a3m file path, default None.
dry_run (bool): Whether to print planned commands without executing design rounds, default False.
skip_env_check (bool): Whether to skip source workflow environment checks, default False.
Return:
status (str): success, error, or partial_success execution status.
msg (str): Human-readable execution summary.
output_dir (str): Unique run directory under tool_result/evobind_tool_result.
fasta (str): Resolved absolute FASTA input path.
peptide_length (int): Effective peptide length used in this run.
num_designs (int): Effective number of design rounds used in this run.
num_iterations (int): Effective number of iterations used in this run.
max_recycles (int): Effective recycle count used in this run.
model_name (str): Effective model name used in this run.
target_residues (str): Effective target residue specification used in this run.
cyclic (bool): Effective cyclic flag used in this run.
dry_run (bool): Effective dry-run flag used in this run.
skip_env_check (bool): Effective environment-check skip flag used in this run.
output_files (dict): Key output file paths including run logs and summary artifacts when available.
metrics (dict): Parsed summary metrics such as candidate count and top-ranked scores when available.
How to use tool evobind_tool :
response = await client.session.call_tool(
"evobind_tool",
arguments={
"fasta": "relative/path/to/receptor.fasta",
"peptide_length": 10,
"num_designs": 10,
"num_iterations": 100,
"max_recycles": 1,
"model_name": "model_1",
"target_residues": "all",
"cyclic": False,
"dry_run": False,
"skip_env_check": False
}
)
result = client.parse_result(response)
key_output = result["output_dir"]
Example parameter sets
{
"fasta": "relative/path/to/1ssc_receptor.fasta",
"peptide_length": 10,
"num_designs": 10,
"num_iterations": 100,
"max_recycles": 1,
"model_name": "model_1",
"target_residues": "all",
"cyclic": False,
"dry_run": False,
"skip_env_check": False
}
{
"fasta": "relative/path/to/target.fasta",
"peptide_length": 12,
"num_designs": 50,
"num_iterations": 500,
"max_recycles": 3,
"model_name": "model_2",
"target_residues": "10,15,20,25",
"cyclic": True,
"msa_file": "relative/path/to/receptor.a3m",
"dry_run": True,
"skip_env_check": True
}
⚠ Mandatory Design Count Verification (L3 Principle 11)
After calling EvoBind, programmatically count the actual number of designs returned. Report the ACTUAL count (not the requested num_designs). If actual < requested, note the discrepancy.
⚠ Mandatory Structure File Download (L3 Principle 14)
Download ALL designed peptide structure files from the output directory. These are Category A files essential for downstream Chai-1 validation and ProLIF analysis.
⚠ ipTM is a Design Score, NOT an Independent Validation (L3 Principle 9)
EvoBind's ipTM is the design tool's own scoring metric. Independent validation with Chai-1 is mandatory before concluding that a design is successful. See L2 Skill 09 for the full validation protocol.
