UniAF3

https://github.com/y1zhou/uniaf3
Admin
UniAF3 is a Python package and CLI tool that provides a unified YAML-based input format to prepare...
Tokens:10,136
Snippets:124
Trust Score:8.9
License:Apache-2.0
Update:2 months ago
Show doc for...
Tokens:
Raw
### Define uniaf3 configuration in YAML

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

A comprehensive example showing the structure for defining sequences, covalent bonds, and various restraints in a YAML configuration file.

```yaml
sequences:
  - polymer_type: protein
    id: [A, B]
    sequence: MVLSPADKTNVKAAWGKVGAHAGEYGAEALERMFLS
    msa_dir: dummy_msa/
    modifications:
      - ccd: HY3
        position: 1
    description: Hemoglobin subunit
  - polymer_type: dna
    id: C
    sequence: GATTACA
  - id: D
    ccd:
      - ATP
  - id: E
    smiles: "CC(=O)OC1C[NH+]2CCC1CC2"
  - id: F
    chai_str: NAG
    description: Example glycan

covalent_bonds:
  - atom1:
      chain_id: B
      residue_idx: 2
      atom_name: CA
      residue_name: V
    atom2:
      chain_id: D
      residue_idx: 1
      atom_name: C04
      residue_name: null

contact_restraints:
  - token1:
      chain_id: A
      residue_idx: 5
      atom_name: CG
      residue_name: P
    token2:
      chain_id: B
      residue_idx: 5
      atom_name: null
      residue_name: P
    max_distance: 8.0
    boltz_enable_force: true

pocket_restraints:
  - binder_chain: D
    max_distance: 6.0
    contact_tokens:
      - chain_id: A
        residue_idx: 10
        atom_name: null
        residue_name: N
      - chain_id: B
        residue_idx: 3
        atom_name: null
        residue_name: L

aux:
  seeds:
    - 42
    - 123
  num_trunk_recycles: 3
  num_diffn_timesteps: 200
  num_diffn_samples: 5
  num_trunk_samples: 1
  boltz_affinity_binder_chain: D
```

--------------------------------

### Multi-entity FASTA format

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Example of a multi-entity FASTA file defining various biological entities including proteins, DNA, ligands, and glycans.

```fasta
>protein|Hemoglobin subunit
MVLSPADKTNVKAAWGKVGAHAGEYGAEALERMFLS
>protein|Modified protein
(HY3)VLSPADKTNVKAAWGKVGAHAGEYGAEALERMFLS
>dna|example DNA
GATTACA
>ligand|CCD example
ATP
>ligand|SMILES example
CC(=O)OC1C[NH+]2CCC1CC2
>glycan|multi-ring glycan
NAG(NAG)(BMA)
```

--------------------------------

### Define molecular restraints in CSV

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Example CSV structure for defining covalent, contact, and pocket restraints.

```csv
restraint_id,chainA,res_idxA,chainB,res_idxB,connection_type,confidence,min_distance_angstrom,max_distance_angstrom,comment
restraint0,A,P5@CG,F,@C1,covalent,1.0,0.0,6.0,protein-glycan bond
restraint1,A,V11,B,L35,contact,1.0,0.0,6.0,interface contact
restraint2,B,,A,A14,pocket,1.0,0.0,8.0,binding pocket
```

--------------------------------

### Inline residue modification

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Example of specifying a residue modification inline within a protein sequence using a parenthesized CCD code.

```text
MVLS(HY3)ADKTNVK
```

--------------------------------

### Convert configuration formats

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Converts a configuration file from a source format to a target format and saves it to the specified output directory.

```bash
uniaf3 convert INPUT_CONFIG_FILE OUTPUT_DIR [PREFIX] [--from-format FORMAT] [--to-format FORMAT]
```

```bash
# UniAF3 → AlphaFold3
uniaf3 convert input.yaml output_dir/ --from-format uniaf3 --to-format alphafold3

# Boltz → Chai-1
uniaf3 convert boltz_input.yaml output_dir/ --from-format boltz --to-format chai

# AF3 → Protenix
uniaf3 convert af3_input.json output_dir/ --from-format alphafold3 --to-format protenix
```

--------------------------------

### Configure MSA inference

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Sets up the inference function to use either a local MSA directory or the ColabFold server.

```python
run_inference(
    fasta_file="input.fasta",
    msa_directory="path/to/msa/",
    use_msa_server=True,
    msa_server_url="https://api.colabfold.com"
)
```

--------------------------------

### Validate a configuration file

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Validates an input configuration file and prints its contents. Supports multiple formats via the --format option.

```bash
uniaf3 validate INPUT_CONFIG_FILE [--format FORMAT]
```

```bash
# Validate a UniAF3 config
uniaf3 validate input.yaml

# Validate a Boltz config
uniaf3 validate boltz_input.yaml --format boltz

# Validate an AlphaFold3 JSON
uniaf3 validate af3_input.json -f alphafold3
```

--------------------------------

### Run Protenix inference via CLI

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/protenix-input-format.md

Execute the inference script using the required JSON input path and configuration parameters.

```bash
python runner/inference.py \
  --input_json_path input.json \
  --dump_dir ./output \
  --seeds 42 \
  --sample_diffusion.N_sample 5
```

--------------------------------

### Define a Ligand input

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-server-input-format.md

Use this structure to define a ligand using its CCD code.

```json
{
  "ligand": {
    "ligand": "CCD_ATP",
    "count": 1
  }
}
```

--------------------------------

### Define ligand inputs

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/protenix-input-format.md

Specify ligands using CCD codes, SMILES strings, or file paths to molecular structure files.

```json
{"ligand": {"ligand": "CCD_ATP", "count": 1}}
{"ligand": {"ligand": "CCD_NAG_BMA_BGC", "count": 1}}
{"ligand": {"ligand": "CC(=O)OC1C[NH+]2CCC1CC2", "count": 1}}
{"ligand": {"ligand": "FILE_/path/to/molecule.sdf", "count": 1}}
```

--------------------------------

### Define Ligand Entry

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Ligands can be defined using either CCD codes or SMILES strings.

```yaml
# CCD ligand (single or multi-CCD)
- id: E
  ccd:
    - ATP

# Multi-CCD ligand (e.g., glycan as ligand)
- id: F
  ccd:
    - NAG
    - BMA

# SMILES ligand
- id: G
  smiles: "CC(=O)OC1C[NH+]2CCC1CC2"
```

--------------------------------

### MSA directory structure

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Expected file naming convention for precomputed A3M files within the MSA directory.

```text
msa_dir/
  a3ms/
    {sha256_hash}.single.a3m
    {sha256_hash}.pair.a3m
```

--------------------------------

### Define Glycan Entry

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Glycans are defined using Chai-1's glycan notation via the chai_str field.

```yaml
- id: H
  chai_str: "NAG(4-1 NAG(4-1 BMA(3-1 MAN)(6-1 MAN)))"
  description: "Branched glycan"
```

--------------------------------

### Define MSA Directory Structure

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Expected directory layout for MSA files, where {seq_hash} is the SHA-256 hex digest of the protein sequence.

```text
msa_dir/
  a3ms/
    {seq_hash}.single.a3m    # Unpaired MSA
    {seq_hash}.pair.a3m      # Paired MSA (optional)
```

--------------------------------

### Define ion ligands

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

Ions are represented as ligands using the standard ligand schema.

```json
{"ligand": {"id": "X", "ccdCodes": ["MG"]}}
```

--------------------------------

### Configure Inference Parameters

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Set optional inference parameters within the aux field, such as recycle counts and diffusion timesteps.

```yaml
aux:
  num_trunk_recycles: 3         # Default: 3
  num_diffn_timesteps: 200      # Default: 200
  num_diffn_samples: 5          # Default: 5
  num_trunk_samples: 1          # Default: 1
  name: "job_name"              # Optional, used in AF3 Server
  boltz_affinity_binder_chain: D  # Boltz-specific: affinity binder chain ID
```

--------------------------------

### Define an Ion input

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-server-input-format.md

Use this structure to define an ion. Note that ion codes do not require the CCD_ prefix.

```json
{
  "ion": {
    "ion": "MG",
    "count": 2
  }
}
```

--------------------------------

### Define Structural Templates

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/boltz-input-format.md

Configures structural templates using CIF or PDB files with optional chain mapping and force enforcement.

```yaml
templates:
  - cif: path/to/template.cif      # or pdb: path/to/template.pdb
    chain_id: A                      # Optional: which chain(s) to template
    template_id: A                   # Optional: explicit template chain mapping
    force: true                      # Optional: enforce template with potential
    threshold: 2.0                   # Optional: max deviation in Å (required if force=true)
```

--------------------------------

### Configure Random Seeds

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Define a list of integer random seeds in the aux.seeds field.

```yaml
aux:
  seeds:
    - 42
    - 123
```

--------------------------------

### Define custom ligand CCD

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

Provide a custom ligand definition in mmCIF format within the userCCD field when specific atom names are required or RDKit conformer generation fails.

```json
{
  "userCCD": "data_MY_LIGAND\n_chem_comp.id MY_LIGAND\n...",
  "sequences": [
    {"ligand": {"id": "A", "ccdCodes": ["MY_LIGAND"]}}
  ]
}
```

--------------------------------

### Define UniAF3 Configuration Structure

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

The top-level YAML structure for UniAF3 configuration, including optional restraints and inference parameters.

```yaml
sequences:
  - # Polymer, Ligand, or Glycan entries
covalent_bonds:   # Optional
  - # CovalentBond entries
contact_restraints:   # Optional
  - # ContactRestraint entries
pocket_restraints:   # Optional
  - # PocketRestraint entries
aux:   # Optional, inference parameters
  seeds:
    - 42
  num_trunk_recycles: 3
  num_diffn_timesteps: 200
  num_diffn_samples: 5
  num_trunk_samples: 1
```

--------------------------------

### Define a Protein Chain input

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-server-input-format.md

Use this structure to define a protein chain, including optional glycans, PTMs, and template usage settings.

```json
{
  "proteinChain": {
    "sequence": "PREACHINGS",
    "count": 2,
    "glycans": [
      {"residues": "NAG(NAG)(BMA)", "position": 8}
    ],
    "modifications": [
      {"ptmType": "CCD_HY3", "ptmPosition": 1}
    ],
    "useStructureTemplate": true,
    "maxTemplateDate": "2024-05-08"
  }
}
```

--------------------------------

### Define a complete Boltz input configuration

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/boltz-input-format.md

This YAML structure defines sequences, covalent bonds, contact constraints, pocket constraints, templates, and affinity properties for a prediction task.

```yaml
version: 1
sequences:
  - protein:
      id: [A, B]
      sequence: MVTPEGNVSLV...
      msa: path/to/msa.csv
      modifications:
        - position: 5
          ccd: HY3
      cyclic: false
  - ligand:
      id: C
      ccd: SAH
constraints:
  - bond:
      atom1: [A, 111, SG]
      atom2: [C, 1, SD]
  - contact:
      token1: [A, 10]
      token2: [C, C4]
      max_distance: 8.0
      force: true
  - pocket:
      binder: C
      contacts:
        - [A, 140]
        - [A, 145]
      max_distance: 6.0
      force: false
templates:
  - cif: path/to/template.cif
    chain_id: A
    force: true
    threshold: 2.0
properties:
  - affinity:
      binder: C
```

--------------------------------

### Define an RNA Sequence input

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-server-input-format.md

Use this structure to define an RNA sequence with optional base modifications.

```json
{
  "rnaSequence": {
    "sequence": "GUAC",
    "count": 1,
    "modifications": [
      {"modificationType": "CCD_2MG", "basePosition": 1}
    ]
  }
}
```

--------------------------------

### Define a Protenix job structure

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/protenix-input-format.md

The top-level structure is a list of job objects containing sequences, covalent bonds, and constraints.

```json
[
  {
    "name": "Test Fold Job",
    "sequences": [...],
    "covalent_bonds": [...],
    "constraint": {...}
  }
]
```

--------------------------------

### Define a Pocket Constraint

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/boltz-input-format.md

Defines a binding pocket by specifying a binder chain and a list of contact points.

```yaml
- pocket:
    binder: CHAIN_ID
    contacts: [[CHAIN_ID, RES_IDX_OR_ATOM_NAME], ...]
    max_distance: 6.0
    force: false
```

--------------------------------

### Define a DNA Sequence input

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-server-input-format.md

Use this structure to define a DNA sequence with optional base modifications.

```json
{
  "dnaSequence": {
    "sequence": "GATTACA",
    "count": 1,
    "modifications": [
      {"modificationType": "CCD_6OG", "basePosition": 1}
    ]
  }
}
```

--------------------------------

### run_inference

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Executes the protein structure prediction inference process using the specified FASTA input and MSA configuration.

```APIDOC
## run_inference

### Description
Executes the protein structure prediction inference process. It supports using a remote MSA server or local precomputed MSA files.

### Parameters
- **fasta_file** (string) - Required - Path to the input FASTA file.
- **msa_directory** (string) - Optional - Path to the directory containing precomputed A3M files.
- **use_msa_server** (boolean) - Optional - Whether to use the ColabFold server for MSA generation.
- **msa_server_url** (string) - Optional - URL of the MSA server to use.
- **num_trunk_recycles** (int) - Optional - Number of recycling steps (default: 3).
- **num_diffn_timesteps** (int) - Optional - Number of diffusion timesteps (default: 200).
- **num_diffn_samples** (int) - Optional - Number of diffusion samples (default: 5).
- **num_trunk_samples** (int) - Optional - Number of trunk samples (default: 1).
- **seed** (int) - Optional - Random seed for reproducibility.
- **use_esm_embeddings** (boolean) - Optional - Whether to use ESM embeddings (default: True).
```

--------------------------------

### Define a multi-ring glycan with bonds

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Specifies a bond between two sugar units using O-glycosidic notation.

```text
NAG(4-1 NAG)
```

--------------------------------

### Define Pocket Constraints

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/protenix-input-format.md

Defines a binding pocket constraint by specifying a binder chain and a list of contact residues. Only one pocket constraint is supported per job.

```json
"constraint": {
  "pocket": {
    "binder_chain": {"entity": 2, "copy": 1},
    "contact_residues": [
      {"entity": 1, "copy": 1, "position": 126}
    ],
    "max_distance": 6.0
  }
}
```

--------------------------------

### Define a DNA sequence in JSON

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

Structure for defining a DNA chain, which does not support MSA configurations.

```json
{
  "dna": {
    "id": "A",
    "sequence": "GACCTCT",
    "modifications": [
      {"modificationType": "6OG", "basePosition": 1}
    ],
    "description": "DNA strand"
  }
}
```

--------------------------------

### Define Protein Sequence Entry

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Schema for protein sequences, supporting MSA directories, structural templates, and PTMs.

```yaml
- polymer_type: protein
  id: A                         # or [A, B] for homomeric copies
  sequence: MVLSPADKTNVK       # Standard 1-letter amino acid codes
  description: "My protein"     # Optional description
  modifications:                # Optional PTMs
    - ccd: HY3                  # CCD code of modification
      position: 1               # 1-based residue index
  msa_dir: path/to/msa/         # Optional, directory containing MSA files
  templates:                    # Optional structural templates
    - path: template.cif        # Path to mmCIF or PDB file
      query_idx: [0, 1, 2]      # 0-based query residue indices
      template_idx: [0, 1, 2]   # 0-based template residue indices
      query_chains: [A]         # Optional, chain IDs in query
      template_chains: [A]      # Optional, chain IDs in template
      boltz_enable_force: false  # Boltz-specific: enforce template
      boltz_template_threshold: null  # Boltz-specific: deviation threshold (Å)
  boltz_cyclic: false           # Boltz-specific: cyclic polymer flag
```

--------------------------------

### Define a Protein sequence in JSON

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

Structure for defining a protein chain, including optional post-translational modifications, MSA paths, and structural templates.

```json
{
  "protein": {
    "id": "A",
    "sequence": "PVLSCGEWQL",
    "modifications": [
      {"ptmType": "HY3", "ptmPosition": 1}
    ],
    "description": "My protein chain",
    "unpairedMsa": "...",
    "unpairedMsaPath": "path/to/unpaired.a3m",
    "pairedMsa": "...",
    "pairedMsaPath": "path/to/paired.a3m",
    "templates": [
      {
        "mmcif": "...",
        "mmcifPath": "path/to/template.cif",
        "queryIndices": [0, 1, 2],
        "templateIndices": [0, 1, 2]
      }
    ]
  }
}
```

--------------------------------

### Define an RNA sequence in JSON

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

Structure for defining an RNA chain, supporting specific modifications and optional unpaired MSA data.

```json
{
  "rna": {
    "id": "A",
    "sequence": "AGCU",
    "modifications": [
      {"modificationType": "2MG", "basePosition": 1}
    ],
    "description": "Short RNA",
    "unpairedMsa": "...",
    "unpairedMsaPath": "..."
  }
}
```

--------------------------------

### Define RNA Sequence Entry

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Schema for RNA sequences, restricted to A, U, G, and C.

```yaml
- polymer_type: rna
  id: D
  sequence: AGCU           # Only A, U, G, C allowed
  modifications:           # Optional
    - ccd: 2MG
      position: 1
```

--------------------------------

### Define Affinity Binder Chain

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/boltz-input-format.md

Specifies the chain ID of the small molecule ligand for affinity computation in the YAML configuration.

```yaml
properties:
  - affinity:
      binder: C    # Chain ID of the small molecule
```

--------------------------------

### Define a single-ring glycan

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Represents a single sugar unit in the Chai-1 notation.

```text
NAG
```

--------------------------------

### Define AlphaFold Server Job Structure

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-server-input-format.md

The top-level structure for an AlphaFold Server request is a JSON array containing one or more job objects.

```json
[
  {
    "name": "Job name",
    "modelSeeds": [],
    "sequences": [...],
    "dialect": "alphafoldserver",
    "version": 1
  }
]
```

--------------------------------

### Define Contact Constraints

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/protenix-input-format.md

Sets soft distance constraints between specific atoms or tokens of two entities.

```json
"constraint": {
  "contact": [
    {
      "entity1": 1, "copy1": 1, "position1": 169,
      "atom1": "CA",
      "entity2": 2, "copy2": 1, "position2": 1,
      "atom2": "C5",
      "max_distance": 6.0,
      "min_distance": 0.0
    }
  ]
}
```

--------------------------------

### Define a Contact Constraint

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/boltz-input-format.md

Sets a distance constraint between two residues or atoms with an optional force parameter.

```yaml
- contact:
    token1: [CHAIN_ID, RES_IDX_OR_ATOM_NAME]
    token2: [CHAIN_ID, RES_IDX_OR_ATOM_NAME]
    max_distance: 6.0
    force: false
```

--------------------------------

### Define DNA Sequence Entry

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Schema for DNA sequences, restricted to A, T, G, and C.

```yaml
- polymer_type: dna
  id: C
  sequence: GATTACA        # Only A, T, G, C allowed
  modifications:           # Optional
    - ccd: 6OG
      position: 1
```

--------------------------------

### Define AlphaFold3 JSON Input Structure

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

The top-level JSON schema for AlphaFold3 input, defining required fields like sequences, modelSeeds, and version.

```json
{
  "name": "Job name",
  "modelSeeds": [1, 2],
  "sequences": [
    {"protein": {...}},
    {"rna": {...}},
    {"dna": {...}},
    {"ligand": {...}}
  ],
  "bondedAtomPairs": [...],
  "userCCD": "...",
  "userCCDPath": "...",
  "dialect": "alphafold3",
  "version": 4
}
```

--------------------------------

### Define a branched glycan

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/chai1-input-format.md

Represents complex branched glycan structures using nested bond notation.

```text
NAG(4-1 NAG(4-1 BMA(3-1 MAN)(6-1 MAN)))
```

--------------------------------

### Define bonded atom pairs

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/alphafold3-input-format.md

Specify covalent bonds between atoms using a list of pairs containing entity ID, residue ID, and atom name.

```json
"bondedAtomPairs": [
  [["A", 1, "CA"], ["F", 1, "CHA"]],
  [["I", 1, "O6"], ["I", 2, "C1"]]
]
```

--------------------------------

### Define a Bond Constraint

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/boltz-input-format.md

Specifies a covalent bond between two atoms using 1-based residue indices and CCD atom names.

```yaml
- bond:
    atom1: [CHAIN_ID, RES_IDX, ATOM_NAME]
    atom2: [CHAIN_ID, RES_IDX, ATOM_NAME]
```

--------------------------------

### Define Covalent Bonds

Source: https://github.com/y1zhou/uniaf3/blob/master/docs/protenix-input-format.md

Specifies covalent bonds between entities using 1-based indexing for entities, copies, and positions.

```json
"covalent_bonds": [
  {
    "entity1": 1,
    "copy1": 1,
    "position1": 2,
    "atom1": "N6",
    "entity2": 3,
    "copy2": 1,
    "position2": 1,
    "atom2": "C1"
  }
]
```

--------------------------------

### Define Pocket Restraints

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Specify a binding pocket where a binder chain interacts with specific contact residues; not supported by AF3 or AF3 Server.

```yaml
pocket_restraints:
  - binder_chain: E          # ID of the chain binding to the pocket
    contact_tokens:           # List of residues forming the pocket
      - chain_id: A
        residue_idx: 10
        atom_name: null       # For polymers; use atom_name for ligands
        residue_name: K
      - chain_id: A
        residue_idx: 15
        atom_name: null
        residue_name: G
    max_distance: 6.0         # Maximum distance in Å (4-20 Å)
    min_distance: 0.0         # Protenix only
    boltz_enable_force: false  # Boltz-specific: enforce with potential
```

--------------------------------

### Define Contact Restraints

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Set distance constraints between two atoms or residues, noting that AF3 and AF3 Server do not support these.

```yaml
contact_restraints:
  - token1:
      chain_id: A
      residue_idx: 10       # 1-based, or 0 if atom_name is used for ligands
      atom_name: null        # Optional for polymers, required for ligands
      residue_name: K        # Optional, for validation
    token2:
      chain_id: C
      residue_idx: 5
      atom_name: null
      residue_name: null
    max_distance: 8.0        # Maximum distance in Å (must be 4-20 Å)
    min_distance: 0.0        # Minimum distance in Å (Protenix only)
    boltz_enable_force: true  # Boltz-specific: enforce with potential
```

--------------------------------

### Define Covalent Bonds

Source: https://github.com/y1zhou/uniaf3/blob/master/README.md

Specify covalent bonds between atoms from different entities using chain IDs and residue indices.

```yaml
covalent_bonds:
  - atom1:
      chain_id: A           # Entity ID
      residue_idx: 5        # 1-based residue index (0 for ligands)
      atom_name: CG         # Atom name (e.g., CA, N, SG)
      residue_name: P       # Optional, for validation
    atom2:
      chain_id: E           # Entity ID
      residue_idx: 1        # 1-based position within ligand
      atom_name: C04        # Atom name in the ligand
      residue_name: null    # Not required for ligands
    description: "Optional description"
```
Tokens:
Raw