### Start AiZynthFinder Application

Source: https://github.com/molecularai/aizynthfinder/blob/master/contrib/notebook.ipynb

Initializes and starts the AiZynthApp with a specified configuration file. This is run after installation to launch the application interface.

```python
#@title Start application. {display-mode: "form"}

from rdkit.Chem.Draw import IPythonConsole
from aizynthfinder.interfaces import AiZynthApp
application = AiZynthApp("./data/config.yml")

```

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

### Install AiZynthFinder for Developers

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Clone the repository, create a development environment using env-dev.yml, and install the package in editable mode with all extras.

```bash
conda env create -f env-dev.yml
conda activate aizynth-dev
poetry install --all-extras
```

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

### Install AiZynthFinder and Dependencies

Source: https://github.com/molecularai/aizynthfinder/blob/master/contrib/notebook.ipynb

Installs the AiZynthFinder package with all optional dependencies, a specific version of Pillow, and downloads public data. This is the initial setup step.

```python
#@title Installation -- Run this cell to install AiZynthFinder

# This might not install the dependencies that you need
#! (AIZ_LATEST_TAG=$(git -c 'versionsort.suffix=-' \
#     ls-remote --exit-code --refs --sort='version:refname' --tags https://github.com/MolecularAI/aizynthfinder '*.*.*' \
#     | tail --lines=1 \
#     | cut -f 3 -d /) \
#   && pip install --quiet https://github.com/MolecularAI/aizynthfinder/archive/${AIZ_LATEST_TAG}.tar.gz)
!pip install --quiet aizynthfinder[all]
!pip install --ignore-installed Pillow==9.0.0
!mkdir --parents data && download_public_data data

```

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

### Get CLI Help

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/cli.rst

Display available command-line arguments and their descriptions for the aizynthcli tool. This is useful for understanding all available options and flags.

```bash
aizynthcli -h
```

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

### Start Jupyter Notebook Server

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/gui.rst

Command to launch the Jupyter Notebook server. This is the first step to access the AIZynthFinder GUI.

```bash
jupyter notebook
```

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

### Use Multiple Expansion Policies with aizynthcli

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Command-line example to utilize the configured multiple expansion policies with `aizynthcli`. Ensure the configuration file (`config.yml`) and input SMILES file (`smiles.txt`) are correctly specified.

```bash
aizynthcli --smiles smiles.txt --config config.yml --policy multi_expansion_strategy
```

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

### Install AiZynthFinder for End-Users

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Create a conda environment and install the AiZynthFinder package with all functionalities.

```bash
conda create "python>=3.10,<3.13" -n aizynth-env
conda activate aizynth-env
python -m pip install aizynthfinder[all]
```

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

### Install AiZynthFinder Minimal Package

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Create a conda environment and install the AiZynthFinder package with essential functionalities.

```bash
conda activate aizynth-env
python -m pip install aizynthfinder
```

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

### Get Help for Molecule Class and Methods

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Access documentation for the Molecule class and its methods, such as 'fingerprint', using the help() function in an interactive Python shell.

```python
from aizynthfinder.chem import Molecule
help(Molecule)
help(Molecule.fingerprint)
```

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

### Configure Retro* Search Algorithm

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Specify Retro* as the search algorithm in the configuration file. This example uses Retro* without a constant-valued oracle function.

```yaml
search:
  algorithm: aizynthfinder.search.retrostar.search_tree.SearchTree
```

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

### Configure Chemformer Expansion Model

Source: https://github.com/molecularai/aizynthfinder/blob/master/plugins/README.md

Example configuration for using the Chemformer expansion model in AIZynthFinder. Ensure the Chemformer REST API is running and accessible.

```yaml
expansion:
    chemformer:
        type: expansion_strategies.ChemformerBasedExpansionStrategy
        url: http://localhost:8000/chemformer-api/predict
search:
    algorithm_config:
        immediate_instantiation: [chemformer]
    time_limit: 300
```

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

### Implement Custom Stock Query

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Example of a custom stock class implementing the __contains__ method for lookup and subclassing StockQueryMixin. This example checks for a heavy atom count less than 10.

```python
from rdkit.Chem import Lipinski
from aizynthfinder.context.stock.queries import StockQueryMixin
class CriteriaStock(StockQueryMixin):
    def __contains__(self, mol):
        return Lipinski.HeavyAtomCount(mol.rd_mol) < 10
```

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

### Configure ModelZoo Expansion Model

Source: https://github.com/molecularai/aizynthfinder/blob/master/plugins/README.md

Example configuration for using the ModelZoo expansion model with Chemformer as the external model. Adjust paths to your cloned repositories and model files.

```yaml
expansion:
    chemformer:
        type: expansion_strategies.ModelZooExpansionStrategy:
        module_path: /path_to_folder_containing_cloned_repository/modelsmatter_modelzoo/external_models/modelsmatter_chemformer_hpc/
        use_gpu: False
        params:
            module_path: /path_to_model_file/chemformer_backward.ckpt
            vocab_path: /path_to_vocab_file/bart_vocab_downstream.txt
search:
    algorithm_config:
        immediate_instantiation: [chemformer]
    time_limit: 300
```

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

### Configure Route Output Limits

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Control the number of routes generated by `aizynthcli` using the `post_processing` section. This example sets a minimum of 25 and a maximum of 50 routes per target.

```yaml
post_processing:
  min_routes: 25
  max_routes: 50
```

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

### Implement Custom Scorer Class

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/scoring.rst

Example of implementing a custom scoring function by inheriting from the Scorer class. This scorer calculates the delta number of transforms.

```python
from aizynthfinder.context.scoring.scorers import Scorer

class DeltaNumberOfTransformsScorer(Scorer):

    def __repr__(self):
        return "delta number of transforms"

    def _score_node(self, node):
        return self._config.max_transforms - node.state.max_transforms

    def _score_reaction_tree(self, tree):
        return self._config.max_transforms - len(list(tree.reactions()))
```

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

### Custom SMILES Extraction Module

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Example Python module to extract SMILES from files with custom formatting, such as Zinc database downloads with headers. This module yields SMILES strings, skipping the first line and splitting subsequent lines by space.

```python
def extract_smiles(filename):
    with open(filename, "r") as fileobj:
        for i, line in enumerate(fileobj.readlines()):
            if i == 0:
                continue
            yield line.strip().split(" ")[0]
```

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

### Create Notebook with aizynthapp tool

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/gui.rst

Command to automatically create a Jupyter notebook and open it. It requires a configuration file path.

```bash
aizynthapp --config config_local.yml
```

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

### Build Documentation with Invoke

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Generate the HTML documentation using Sphinx via an invoke command.

```bash
invoke build-docs
```

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

### Initialize Clustering GUI

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/gui.rst

Enables the clustering GUI extension for analyzing search results. Requires the matplotlib backend to be set inline.

```python
%matplotlib inline
from aizynthfinder.interfaces.gui.clustering import ClusteringGui
ClusteringGui.from_app(app)
```

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

### Simple AI-ZynthFinder Configuration

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/configuration.rst

Basic configuration for expansion models and stock files. Ensure the specified model and template files exist.

```yaml
expansion:
  full:
    - uspto_expansion.onnx
    - uspto_templates.csv.gz
stock:
  zinc: zinc_stock.hdf5
```

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

### Run Full Tests with Invoke

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Execute the full test suite as run on the CI server using an invoke command.

```bash
invoke full-tests
```

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

### Instantiate AiZynthExpander

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Instantiate the AiZynthExpander class with a configuration file to break down molecules into reactants.

```python
filename = "config.yml"
expander = AiZynthExpander(configfile=filename)
expander.expansion_policy.select("uspto")
expander.filter_policy.select("uspto")
```

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

### Instantiate AiZynthFinder

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Instantiate the AiZynthFinder class by providing the path to a configuration file.

```python
filename = "config.yml"
finder = AiZynthFinder(configfile=filename)
```

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

### Load Custom Stock in AiZynthApp

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Demonstrates how to load a custom stock class within the AiZynthApp. The custom stock is instantiated and then loaded into the finder's stock object.

```python
from aizynthfinder import AiZynthApp
configfile="config_local.yml"
app = AiZynthApp(configfile, setup=False)
app.finder.stock.load(CriteriaStock(), "criteria") # This loads the custom stock class
app.setup()
```

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

### Initialize AiZynthApp in Jupyter

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/index.rst

Instantiates the AiZynthApp for the GUI interface within a Jupyter notebook. Requires the path to a configuration file.

```python
from aizynthfinder.interfaces import AiZynthApp
app = AiZynthApp("/path/to/configfile.yaml")
```

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

### AI-ZynthFinder Configuration with Environment Variables

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/configuration.rst

Configuration using environment variables for search parameters. Ensure the corresponding environment variables are set before running.

```yaml
search:
  iteration_limit: ${ITERATION_LIMIT}
  algorithm_config:
    C: ${C}
  time_limit: ${TIME_LIMIT}
  max_transforms: ${MAX_TRANSFORMS}
```

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

### Generate Images of Reaction Routes

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/cli.rst

Create PNG images for the top-ranked reaction routes of the first target compound from the analysis data. This requires the pandas and ReactionTree modules.

```python
import pandas as pd
from aizynthfinder.reactiontree import ReactionTree

data = pd.read_json("output.json.gz", orient="table")
all_trees = data.trees.values  # This contains a list of all the trees for all the compounds
trees_for_first_target = all_trees[0]

for itree, tree in enumerate(trees_for_first_target):
    imagefile = f"route{itree:03d}.png"
    ReactionTree.from_dict(tree).to_image().save(imagefile)
```

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

### Create MongoDB Stock from SMILES

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Command-line tool to create a MongoDB stock database from one or more SMILES files. Specify a target name for the database.

```bash
smiles2stock --files file1.smi file2.smi --output my_db --target mongo
```

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

### Create HDF5 Stock from SMILES

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Command-line tool to create an HDF5 stock file from one or more SMILES files. Each line in the input files should contain a single SMILES string.

```bash
smiles2stock --files file1.smi file2.smi --output stock.hdf5
```

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

### Configure Retro* with Oracle Function

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Configure Retro* to use a specific oracle function by providing the molecule cost details, including the model path and fingerprint parameters.

```yaml
search:
  algorithm: aizynthfinder.search.retrostar.search_tree.SearchTree
  algorithm_config:
    molecule_cost: 
      cost: aizynthfinder.search.retrostar.cost.RetroStarCost
      model_path: retrostar_value_model.pickle
      fingerprint_length: 2048
      fingerprint_radius: 2
      dropout_rate: 0.1
```

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

### Select Stock and Policies

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Select the desired stock and policies (expansion and filter) for the AiZynthFinder instance. The keys 'zinc' and 'uspto' must correspond to entries in the configuration file.

```python
finder.stock.select("zinc")
finder.expansion_policy.select("uspto")
finder.filter_policy.select("uspto")
```

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

### Configure All Routes Output

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Configure `aizynthcli` to extract all solved routes, falling back to `min_routes` and `max_routes` if a target is unsolved. This ensures maximum data retrieval for solvable targets.

```yaml
post_processing:
    min_routes: 5
    max_routes: 10
    all_routes: True
```

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

### Import AiZynthFinder Class

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Import the main AiZynthFinder class from the aizynthfinder library.

```python
from aizynthfinder.aizynthfinder import AiZynthFinder
```

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

### Run AiZynthFinder in Batch Mode

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/index.rst

Executes the aizynthcli script for batch-mode tree search. Requires a configuration file and a text file with SMILES strings.

```bash
aizynthcli --config config.yml --smiles smiles.txt
```

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

### Download Public Data

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/index.rst

Downloads public data for AiZynthFinder. This data includes a trained model and a stock collection.

```bash
download_public_data .
```

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

### Configure Multiple Expansion Policies

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Enable the use of multiple expansion policies, such as combining a general USPTO model with a RingBreaker model. The `multi_expansion_strategy` defines how these policies are combined.

```yaml
expansion:
  uspto:
    - uspto_keras_model.hdf5
    - uspto_unique_templates.csv.gz
  ringbreaker:
    - uspto_ringbreaker_keras_model.hdf5
    - uspto_ringbreaker_unique_templates.csv.gz
  multi_expansion_strategy:
    type: aizynthfinder.context.policy.MultiExpansionStrategy
    expansion_strategies: [uspto, ringbreaker]
    additive_expansion: True
```

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

### Run Batch Tree Search

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/cli.rst

Execute a batch tree search using a configuration file and a SMILES file. Ensure your configuration file specifies paths to policy models and stocks, and the SMILES file contains one SMILES per line.

```bash
aizynthcli --config config_local.yml --smiles smiles.txt
```

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

### Download Public Data

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Download necessary data files (stock, policy networks) for AiZynthFinder to a specified folder and generate a config.yml.

```bash
download_public_data my_folder
```

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

### Advanced AI-ZynthFinder Configuration

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/configuration.rst

Comprehensive configuration including MCTS search parameters, custom expansion policies, filter models, and stock definitions. Paths to models and templates must be valid.

```yaml
search:
  algorithm: mcts
  algorithm_config:
    C: 1.4
    default_prior: 0.5
    use_prior: True
    prune_cycles_in_search: True
    search_rewards:
      - state score
  max_transforms: 6
  iteration_limit: 100
  return_first: false
  time_limit: 120
  exclude_target_from_stock: True
  break_bonds: [[1, 2], [2, 3]]
  freeze_bonds: [[3, 4]]
  break_bonds_operator: or
expansion:
  my_policy:
    type: template-based
    model: /path/to/keras/model/weights.hdf5
    template: /path/to/hdf5/templates.hdf5
    template_column: retro_template
    cutoff_cumulative: 0.995
    cutoff_number: 50
    use_rdchiral: True
    use_remote_models: False
  my_full_policy:
    - /path/to/keras/model/weights.hdf5
    - /path/to/hdf5/templates.hdf5
filter:
  uspto:
    type: quick-filter
    model: /path/to/keras/model/weights.hdf5
    exclude_from_policy: rc
    filter_cutoff: 0.05
    use_remote_models: False
  uspto_full: /path/to/keras/model/weights.hdf5
stock:
  buyables:
    type: inchiset
    path: /path/to/stock1.hdf5
  emolecules: /path/to/stock1.hdf5
```

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

### Create Pandas DataFrame from Reaction Metadata

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Collect metadata from all reactions in the expansion results and organize it into a pandas DataFrame.

```python
import pandas as pd
metadata = []
for reaction_tuple in reactions:
    for reaction in reaction_tuple:
        metadata.append(reaction.metadata)
df = pd.DataFrame(metadata)
```

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

### Use Custom SMILES Extractor with smiles2stock

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Command to create a stock using a custom SMILES extraction module. Ensure the module is in a Python path and specify it using the --source module option.

```bash
export PYTHONPATH=`pwd`
smiles2stock --files load_zinc file1.smi file2.smi --source module --output stock.hdf5
```

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

### Run AiZynthFinder with Single SMILES

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/index.rst

Executes the aizynthcli script for batch-mode tree search on a single molecule specified directly on the command line.

```bash
aizynthcli --config config.yml --smiles "COc1cccc(OC(=O)/C=C/c2cc(OC)c(OC)c(OC)c2)c1"
```

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

### Configure Disconnection-Aware Chemformer Expansion

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Set up a multi-expansion policy that includes a disconnection-aware Chemformer strategy. This requires specifying the Chemformer API URL and other parameters like `n_beams` and `cutoff_number`.

```yaml
expansion:
    standard:
      type: template-based
      model: path/to/model
      template: path/to/templates
    chemformer_disconnect:
      type: expansion_strategies.DisconnectionAwareExpansionStrategy
      url: "http://localhost:8023/chemformer-disconnect-api/predict-disconnection"
      n_beams: 5
    multi_expansion:
      type: aizynthfinder.context.policy.MultiExpansionStrategy
      expansion_strategies: [chemformer_disconnect, standard]
      additive_expansion: True
      cutoff_number: 50
```

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

### Configure MongoDB Stock

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Specify MongoDB connection details and collection for compound lookups. Defaults to localhost, stock_db, and molecules if no options are provided.

```yaml
stock:
    type: mongodb
    host: user@myurl.com
    database: database_name
    collection: compounds
```

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

### Build Routes and Extract Statistics

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

After the tree search, build the possible routes and extract statistical information about the search results. The build_routes method must be called before analysis.

```python
finder.build_routes()
stats = finder.extract_statistics()
```

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

### Specify Custom Stock in Configuration

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Configure AiZynthFinder to use a custom stock class directly from the configuration file. The class must be located in a Python module known to the interpreter.

```yaml
stock:
    type: aizynthfinder.contrib.stocks.CriteriaStock
```

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

### Add Custom Scorer to AiZynthApp

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/scoring.rst

Demonstrates how to add a custom scorer to an AiZynthApp instance. This involves creating the scorer object and loading it into the finder's scorers collection.

```python
from aizynthfinder.interfaces import AiZynthApp
app = AiZynthApp("config_local.yml", setup=False)
scorer = DeltaNumberOfTransformsScorer(app.finder.config)
app.finder.scorers.load(scorer)
app.setup()
```

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

### Run Pytest Tests

Source: https://github.com/molecularai/aizynthfinder/blob/master/README.md

Execute all tests using the pytest package.

```bash
pytest -v
```

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

### Read Output JSON

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/cli.rst

Load the default output JSON file generated by aizynthcli into a pandas DataFrame for analysis. This file typically contains statistics and routes for multiple target compounds.

```python
import pandas as pd
data = pd.read_json("output.json.gz", orient="table")
```

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

### Configure Stop Criteria

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/stocks.rst

Define stop criteria for tree search based on price and element counts. Requires the stock query class to support price and amount querying.

```yaml
stock:
    stop_criteria:
        price: 10
        counts:
            C: 10
```

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

### Perform Expansion

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Perform molecular expansion using the configured expander and a target SMILES string. The filter policy is optional and affects feasibility, not filtering.

```python
reactions = expander.do_expansion("Cc1cccc(c1N(CC(=O)Nc2ccc(cc2)c3ncon3)C(=O)C4CCS(=O)(=O)CC4)C")
```

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

### Set Target SMILES and Perform Tree Search

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Set the target molecule using its SMILES string and initiate the tree search process.

```python
finder.target_smiles = "Cc1cccc(c1N(CC(=O)Nc2ccc(cc2)c3ncon3)C(=O)C4CCS(=O)(=O)CC4)C"
finder.tree_search()
```

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

### Configure Bond Constraints for MO-MCTS

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Specify bond constraints for disconnection-aware retrosynthesis. This involves defining `break_bonds` and `freeze_bonds` under the `search` configuration, and including 'broken bonds' in `search_rewards`.

```yaml
search:
      break_bonds: [[1, 2], [3, 4]]
      freeze_bonds: []
      algorithm_config:
        search_rewards: ["state score", "broken bonds"]
```

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

### Extract Reactants SMILES from Reactions

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/python_interface.rst

Iterate through the results of an expansion to extract the SMILES strings of the reactants from each reaction.

```python
reactants_smiles = []
for reaction_tuple in reactions:
    reactants_smiles.append([mol.smiles for mol in reaction_tuple[0].reactants[0])
```

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

### Configure Route Scorers for MO-Tree Ranking

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/howto.rst

Specify route scorers for multi-objective tree ranking and building. This configuration ensures that routes are evaluated based on multiple criteria, such as 'state score' and 'broken bonds'.

```yaml
post_processing:
    route_scorers: ["state score", "broken bonds"]
```

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

### Access Finder and Extract Statistics

Source: https://github.com/molecularai/aizynthfinder/blob/master/docs/gui.rst

Retrieves the finder object from the app and extracts search statistics. This is used after a tree search has been completed.

```python
finder = app.finder
stats = finder.extract_statistics()
```

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