### Download RCSB Data and Start Redis

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Download the CCD RDB file and start a Redis server. Ensure Redis is installed and accessible.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/ccd.rdb
redis-server --dbfilename ccd.rdb --port 7777
```

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

### Install Project Requirements

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Installs Python dependencies from the requirements.txt file. Ensure you have pip installed.

```bash
pip install -r requirements.txt
```

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

### Configure Multiple Dataset Sources

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Example configuration for setting up multiple dataset sources, such as PDB and OpenFold distillation datasets, for training the structure model. Each dataset can have its own probability, sampler, cropper, and split configuration.

```yaml
  datasets:
    - _target_: boltz.data.module.training.DatasetConfig
      target_dir: PATH_TO_PDB_TARGETS_DIR
      msa_dir: PATH_TO_PDB_MSA_DIR
      prob: 0.5
      sampler:
        _target_: boltz.data.sample.cluster.ClusterSampler
      cropper:
        _target_: boltz.data.crop.boltz.BoltzCropper
        min_neighborhood: 0
        max_neighborhood: 40
      split: ./scripts/train/assets/validation_ids.txt
    - _target_: boltz.data.module.training.DatasetConfig
      target_dir: PATH_TO_DISTILLATION_TARGETS_DIR
      msa_dir: PATH_TO_DISTILLATION_MSA_DIR
      prob: 0.5
      sampler:
        _target_: boltz.data.sample.cluster.ClusterSampler
      cropper:
        _target_: boltz.data.crop.boltz.BoltzCropper
        min_neighborhood: 0
        max_neighborhood: 40
```

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

### Install Boltz from GitHub

Source: https://github.com/jwohlwend/boltz/blob/main/README.md

Install the latest development version of Boltz directly from its GitHub repository. This method is useful for accessing daily updates.

```bash
git clone https://github.com/jwohlwend/boltz.git
cd boltz; pip install -e .[cuda]
```

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

### Launch Redis Server for MSA Processing

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Starts a Redis server instance configured to use the downloaded taxonomy database. This is used for parallel MSA processing.

```bash
redis-server --dbfilename taxonomy.rdb --port 7777
```

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

### Install Boltz with CUDA Support

Source: https://github.com/jwohlwend/boltz/blob/main/README.md

Install the Boltz package with CUDA acceleration using pip. This is the recommended method for GPU users.

```bash
pip install boltz[cuda] -U
```

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

### FASTA Format Example

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Example of input entries in the deprecated FASTA format, including chain ID, entity type, and MSA path or sequence.

```plaintext
>A|protein|./examples/msa/seq1.a3m
MVTPEGNVSLVDESLLVGVTDEDRAVRSAHQFYERLIGLWAPAVMEAAHELGVFAALAEAPADSGELARRLDCDARAMRVLLDALYAYDVIDRIHDTNGFRYLLSAEARECLLPGTLFSLVGKFMHDINVAWPAWRNLAEVVRHGARDTSGAESPNGIAQEDYESLVGGINFWAPPIVTTLSRKLRASGRSGDATASVLDVGCGTGLYSQLLLREFPRWTATGLDVERIATLANAQALRLGVEERFATRAGDFWRGGWGTGYDLVLFANIFHLQTPASAVRLMRHAAACLAPDGLVAVVDQIVDADREPKTPQDRFALLFAASMTNTGGGDAYTFQEYEEWFTAAGLQRIETLDTPMHRILLARRATEPSAVPEGQASENLYFQ
>B|protein|./examples/msa/seq1.a3m
MVTPEGNVSLVDESLLVGVTDEDRAVRSAHQFYERLIGLWAPAVMEAAHELGVFAALAEAPADSGELARRLDCDARAMRVLLDALYAYDVIDRIHDTNGFRYLLSAEARECLLPGTLFSLVGKFMHDINVAWPAWRNLAEVVRHGARDTSGAESPNGIAQEDYESLVGGINFWAPPIVTTLSRKLRASGRSGDATASVLDVGCGTGLYSQLLLREFPRWTATGLDVERIATLANAQALRLGVEERFATRAGDFWRGGWGTGYDLVLFANIFHLQTPASAVRLMRHAAACLAPDGLVAVVDQIVDADREPKTPQDRFALLFAASMTNTGGGDAYTFQEYEEWFTAAGLQRIETLDTPMHRILLARRATEPSAVPEGQASENLYFQ
>C|ccd
SAH
>D|ccd
SAH
>E|smiles
N[C@@H](Cc1ccc(O)cc1)C(=O)O
>F|smiles
N[C@@H](Cc1ccc(O)cc1)C(=O)O
```

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

### Example Prediction Configuration

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

This YAML configuration specifies sequences for protein and ligands, including MSA path and SMILES notation. It serves as a foundational example for setting up a prediction task in Boltz.

```yaml
version: 1
sequences:
  - protein:
      id: [A, B]
      sequence: MVTPEGNVSLVDESLLVGVTDEDRAVRSAHQFYERLIGLWAPAVMEAAHELGVFAALAEAPADSGELARRLDCDARAMRVLLDALYAYDVIDRIHDTNGFRYLLSAEARECLLPGTLFSLVGKFMHDINVAWPAWRNLAEVVRHGARDTSGAESPNGIAQEDYESLVGGINFWAPPIVTTLSRKLRASGRSGDATASVLDVGCGTGLYSQLLLREFPRWTATGLDVERIATLANAQALRLGVEERFATRAGDFWRGGWGTGYDLVLFANIFHLQTPASAVRLMRHAAACLAPDGLVAVVDQIVDADREPKTPQDRFALLFAASMTNTGGGDAYTFQEYEEWFTAAGLQRIETLDTPMHRILLARRATEPSAVPEGQASENLYFQ
      msa: ./examples/msa/seq1.a3m
  - ligand:
      id: [C, D]
      ccd: SAH
  - ligand:
      id: [E, F]
      smiles: 'N[C@@H](Cc1ccc(O)cc1)C(=O)O'
```

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

### Run Training Script (Full)

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Launch the full training run using the specified configuration file. This command assumes debug mode has been tested and is working correctly.

```bash
python scripts/train/train.py scripts/train/configs/structure.yaml
```

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

### Basic Authentication using CLI Options

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Provide username and password directly as command-line arguments for basic authentication with the MSA server.

```bash
boltz predict ... --use_msa_server --msa_server_username myuser --msa_server_password mypassword
```

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

### Download Symmetry Files

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Download the pre-computed symmetry information for ligands, which is provided as a pickle file.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/symmetry.pkl
```

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

### Basic Authentication using Environment Variables

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Set username and password via environment variables for basic authentication with the MSA server. Recommended for security.

```bash
export BOLTZ_MSA_USERNAME=myuser
export BOLTZ_MSA_PASSWORD=mypassword
boltz predict ... --use_msa_server
```

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

### API Key Authentication using CLI Options

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Specify API key header and value using command-line options for API key authentication with the MSA server.

```bash
boltz predict ... --use_msa_server --api_key_header X-API-Key --api_key_value <your-api-key>
```

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

### Run Training Script (Confidence Model)

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Execute the training script using the configuration file specifically designed for training the confidence model.

```bash
python scripts/train/train.py scripts/train/configs/confidence.yaml
```

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

### Download CCD Dictionary

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Downloads the preprocessed CCD dictionary file. This is a necessary step if you are not recreating it yourself.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/ccd.pkl
```

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

### Navigate to Processing Folder

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Change the current directory to the 'scripts/process' folder to begin the data processing pipeline.

```bash
cd scripts/process
```

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

### Download Pre-processed RCSB (PDB) Structures

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Use wget to download the compressed PDB structures, then extract them and remove the archive. Ensure you have sufficient storage.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/rcsb_processed_targets.tar
tar -xf rcsb_processed_targets.tar
rm rcsb_processed_targets.tar
```

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

### Configure Training Parameters (Structure Model)

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Modify the training configuration file (e.g., structure.yaml) to specify data paths, output directories, and model/training parameters. Adjust max_tokens and max_atoms based on GPU size and desired training speed.

```yaml
trainer:
  devices: 1

output: SET_PATH_HERE                 # Path to the output directory  
resume: PATH_TO_CHECKPOINT_FILE       # Path to a checkpoint file to resume training from if any null otherwise

data:
  datasets:
    - _target_: boltz.data.module.training.DatasetConfig
      target_dir: PATH_TO_TARGETS_DIR       # Path to the directory containing the processed structure files
      msa_dir: PATH_TO_MSA_DIR              # Path to the directory containing the processed MSA files

  symmetries: PATH_TO_SYMMETRY_FILE      # Path to the file containing molecule the symmetry information
  max_tokens: 512                        # Maximum number of tokens in the input sequence
  max_atoms: 4608                        # Maximum number of atoms in the input structure
```

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

### Download Taxonomy Database

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Downloads the taxonomy database file required for MSA processing. Place it in the current folder.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/taxonomy.rdb
```

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

### Create Sequence Clusters

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Runs the clustering script to assign proteins, RNA/DNA, and ligands to similarity clusters. Requires the CCD dictionary, sequence data, and the path to the mmseqs executable.

```bash
python cluster.py --ccd ccd.pkl --sequences pdb_seqres.txt --mmseqs PATH_TO_MMSEQS_EXECUTABLE --outdir ./clustering
```

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

### Predict Structure with Custom Recycling and Sampling Steps

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Configure the number of recycling steps and diffusion samples for prediction. Note that increasing these values will significantly increase prediction time.

```bash
boltz predict input_path --recycling_steps 10 --diffusion_samples 25
```

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

### API Key Authentication using Environment Variable

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Set the API key value via an environment variable for API key authentication with the MSA server. The default header is X-API-Key.

```bash
export MSA_API_KEY_VALUE=<your-api-key>
boltz predict ... --use_msa_server --api_key_header X-API-Key
```

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

### Download and Extract OpenFold MSAs

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Downloads a large archive of raw OpenFold MSAs and extracts them. Ensure sufficient disk space.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/openfold_raw_msa.tar
tar -xf openfold_raw_msa.tar
rm openfold_raw_msa.tar
```

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

### Run Training Script (Debug Mode)

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Execute the training script with the debug flag enabled. This disables DDP, sets num_workers to 0 for single-process execution, and disables wandb for testing.

```bash
python scripts/train/train.py scripts/train/configs/structure.yaml debug=1
```

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

### Download Pre-processed OpenFold Structures

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Download the processed structure files from OpenFold using wget, extract the tar archive, and clean up by removing the tar file.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/openfold_processed_targets.tar
tar -xf openfold_processed_targets.tar
rm openfold_processed_targets.tar
```

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

### Download Pre-processed RCSB (PDB) MSAs

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Download the Multiple Sequence Alignments (MSAs) for PDB structures using wget, extract the archive, and then remove the tar file.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/rcsb_processed_msa.tar
tar -xf rcsb_processed_msa.tar
rm rcsb_processed_msa.tar
```

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

### Recreate CCD Dictionary

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Recreates the CCD dictionary from the PDB components CIF file. Requires the ccd.py script.

```bash
wget https://files.wwpdb.org/pub/pdb/data/monomers/components.cif
python ccd.py --components components.cif --outdir ./ccd
```

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

### Predict Structure with MSA Server

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Use this option to automatically generate the MSA with the mmseqs2 server.

```bash
boltz predict input_path --use_msa_server
```

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

### Download Pre-processed OpenFold MSAs

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Obtain the pre-processed MSAs from OpenFold by downloading the tar archive, extracting its contents, and then deleting the archive.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/openfold_processed_msa.tar
tar -xf openfold_processed_msa.tar
rm openfold_processed_msa.tar
```

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

### Download PDB Sequence Data

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Downloads and unzips the PDB sequence data. This is used as input for the clustering script.

```bash
wget https://files.rcsb.org/pub/pdb/derived_data/pdb_seqres.txt.gz
gunzip -d pdb_seqres.txt.gz
```

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

### Predict Structure with Inference Time Potential

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Enables the use of an inference time potential to improve the physical quality of poses.

```bash
boltz predict input_path --use_potentials
```

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

### Process Raw MSAs

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Runs the MSA processing script. Requires the directory containing raw MSAs, an output directory, and the Redis port. Ensure MSA files have .a3m or .a3m.gz extensions.

```bash
python msa.py --msadir YOUR_MSA_DIR --outdir YOUR_OUTPUT_DIR --redis-port 7777
```

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

### Boltz-1 Evaluation File Structure

Source: https://github.com/jwohlwend/boltz/blob/main/docs/evaluation.md

This structure outlines the organization of input, target, output, and evaluation files for Boltz-1, Boltz-1x, Chai-1, and AlphaFold3. It includes summary CSV files for evaluation results.

```bash
boltz_results_final/
├── inputs/                 # Input files for every model
    ├── casp15/
        ├── af3
        ├── boltz
        ├── boltzx
        ├── chai
        └── msa
    └── test/
├── targets/                # Target files from PDB
    ├── casp15
    └── test
├── outputs/                # Output files for every model
    ├── casp15
    └── test
├── evals/                  # Output of evaluation script for every
    ├── casp15
    └── test
├── results_casp.csv        # Summary of evaluation results for CASP15
└── results_test.csv        # Summary of evaluation results for test set
```

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

### Download and Extract PDB MSAs

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Downloads a large archive of raw PDB Multiple Sequence Alignments (MSAs) and extracts them. Ensure sufficient disk space.

```bash
wget https://boltz1.s3.us-east-2.amazonaws.com/rcsb_raw_msa.tar
tar -xf rcsb_raw_msa.tar
rm rcsb_raw_msa.tar
```

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

### Run Boltz Prediction

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

Execute a prediction using Boltz. Specify the input path and any desired options. Use --override to force re-computation.

```bash
boltz predict <INPUT_PATH> [OPTIONS]
```

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

### Run Boltz Inference

Source: https://github.com/jwohlwend/boltz/blob/main/README.md

Execute the Boltz prediction command. Specify the input path for biomolecule descriptions and desired properties. Use --use_msa_server for remote MSA processing.

```bash
boltz predict input_path --use_msa_server
```

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

### Process PDB Data with rcsb.py

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Run the `rcsb.py` script to process PDB data. Specify the directory containing mmcif files, the clustering configuration, output directory, and Redis port. The script processes `.cif` or `cif.gz` files and can filter by file size.

```bash
python rcsb.py --datadir PATH_TO_MMCIF_DIR --cluster clustering/clustering.json --outdir YOUR_OUTPUT_DIR --use-assembly --max-file-size 7000000 --redis-port 7777
```

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

### Boltz YAML Input Schema

Source: https://github.com/jwohlwend/boltz/blob/main/docs/prediction.md

The complete schema for defining complexes in YAML format for Boltz predictions. Includes sequences, constraints, templates, and properties.

```yaml
sequences:
    - ENTITY_TYPE:
        id: CHAIN_ID 
        sequence: SEQUENCE      # only for protein, dna, rna
        smiles: 'SMILES'        # only for ligand, exclusive with ccd
        ccd: CCD                # only for ligand, exclusive with smiles
        msa: MSA_PATH           # only for protein
        modifications:
          - position: RES_IDX   # index of residue, starting from 1
            ccd: CCD            # CCD code of the modified residue
        cyclic: false
    - ENTITY_TYPE:
        id: [CHAIN_ID, CHAIN_ID]    # multiple ids in case of multiple identical entities
        ...
constraints:
    - bond:
        atom1: [CHAIN_ID, RES_IDX, ATOM_NAME]
        atom2: [CHAIN_ID, RES_IDX, ATOM_NAME]
    - pocket:
        binder: CHAIN_ID
        contacts: [[CHAIN_ID, RES_IDX/ATOM_NAME], [CHAIN_ID, RES_IDX/ATOM_NAME]]
        max_distance: DIST_ANGSTROM
        force: false # if force is set to true (default is false), a potential will be used to enforce the pocket constraint
    - contact:
        token1: [CHAIN_ID, RES_IDX/ATOM_NAME]
        token2: [CHAIN_ID, RES_IDX/ATOM_NAME]
        max_distance: DIST_ANGSTROM
        force: false # if force is set to true (default is false), a potential will be used to enforce the contact constraint

templates:
    - cif: CIF_PATH  # if only a path is provided, Boltz will find the best matchings
      force: true # optional, if force is set to true (default is false), a potential will be used to enforce the template
      threshold: DISTANCE_THRESHOLD # optional, controls the distance (in Angstroms) that the prediction can deviate from the template
    - cif: CIF_PATH
      chain_id: CHAIN_ID   # optional, specify which chain to find a template for
    - cif: CIF_PATH
      chain_id: [CHAIN_ID, CHAIN_ID]  # can be more than one
      template_id: [TEMPLATE_CHAIN_ID, TEMPLATE_CHAIN_ID]
    - pdb: PDB_PATH # if a pdb path is provided, Boltz will incrementally assign template chain ids based on the chain names in the PDB file (A1, A2, B1, etc)
      chain_id: [CHAIN_ID, CHAIN_ID]
      template_id: [TEMPLATE_CHAIN_ID, TEMPLATE_CHAIN_ID]
properties:
    - affinity:
        binder: CHAIN_ID

```

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

### Hash Sequence Function

Source: https://github.com/jwohlwend/boltz/blob/main/docs/training.md

Python function to generate a SHA256 hash for a given sequence string. Used for naming MSA files.

```python
import hashlib

def hash_sequence(seq: str) -> str:
    """Hash a sequence."""
    return hashlib.sha256(seq.encode()).hexdigest()
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.