aizynthfinder.search.retrostar package

Submodules

aizynthfinder.search.retrostar.cost module

Module containing Retro* cost models

class aizynthfinder.search.retrostar.cost.MoleculeCost(config)

Bases: object

A class to compute the molecule cost.

The cost to be computed is taken from the input config. If no molecule_cost is set, assigns ZeroMoleculeCost as the cost by default. The molecule_cost can be set as a dictionary in config under search in the following format: ‘algorithm’: ‘retrostar’ ‘algorithm_config’: {

‘molecule_cost’: {

‘cost’: name of the search cost class or custom_package.custom_model.CustomClass, other settings or params

}

}

The cost can be computed by calling the instantiated class with a molecule.

calculator = MyCost(config)
cost = calculator.calculate(molecule)
Parameters

config (Configuration) – the configuration of the tree search

Return type

None

class aizynthfinder.search.retrostar.cost.RetroStarCost(**kwargs)

Bases: object

Encapsulation of the original Retro* molecular cost model

Numpy implementation of original pytorch model

The predictions of the score is made on a Molecule object

mol = Molecule(smiles="CCC")
scorer = RetroStarCost()
score = scorer.calculate(mol)

The model provided when creating the scorer object should be a pickled tuple. The first item of the tuple should be a list of the model weights for each layer. The second item of the tuple should be a list of the model biases for each layer.

Parameters

  • model_path – the filename of the model weights and biases

  • fingerprint_length – the number of bits in the fingerprint

  • fingerprint_radius – the radius of the fingerprint

  • dropout_rate – the dropout_rate

  • kwargs (Any) –

Return type

None

calculate(mol)
Parameters

mol (Molecule) –

Return type

float

class aizynthfinder.search.retrostar.cost.ZeroMoleculeCost

Bases: object

Encapsulation of a Zero cost model

calculate(mol)
Parameters

mol (Molecule) –

Return type

float

aizynthfinder.search.retrostar.nodes module

Module containing a classes representation various tree nodes

class aizynthfinder.search.retrostar.nodes.MoleculeNode(mol, config, molecule_cost, parent=None)

Bases: aizynthfinder.search.andor_trees.TreeNodeMixin

An OR node representing a molecule

Variables

  • cost – the cost of synthesizing the molecule

  • expandable – if True, this node is part of the frontier

  • mol – the molecule represented by the node

  • in_stock – if True the molecule is in stock and hence should not be expanded

  • parent – the parent of the node

  • solved – if True the molecule is in stock or at least one child node is solved

  • value – the current rn(m|T)

Parameters
Return type

None

classmethod create_root(smiles, config, molecule_cost)

Create a root node for a tree using a SMILES.

Parameters

  • smiles (str) – the SMILES representation of the root state

  • config (Configuration) – settings of the tree search algorithm

  • molecule_cost (MoleculeCost) –

Returns

the created node

Return type

MoleculeNode

classmethod from_dict(dict_, config, molecules, molecule_cost, parent=None)

Create a new node from a dictionary, i.e. deserialization

Parameters
Returns

a deserialized node

Return type

MoleculeNode

property children: List[ReactionNode]

Gives the reaction children nodes

property target_value: float

The V_t(m|T) value, the current cost of the tree containing this node

Returns

the value

property prop: StrDict

Dictionary with publicly exposed properties

add_stub(cost, reaction)

Add a stub / sub-tree to this node

Parameters

  • cost (float) – the cost of the reaction

  • reaction (RetroReaction) – the reaction creating the stub

Returns

list of all newly added molecular nodes

Return type

Sequence[MoleculeNode]

ancestors()

Return the ancestors of this node

Returns

the ancestors

Return type

set

close()

Updates the values of this node after expanding it.

Returns

the delta V value

Return type

float

serialize(molecule_store)

Serialize the node object to a dictionary

Parameters

molecule_store (MoleculeSerializer) – the serialized molecules

Returns

the serialized node

Return type

StrDict

update(solved)

Update the node as part of the update algorithm, calling the update() method of its parent if available.

Parameters

solved (bool) – if the child node was solved

Return type

None

class aizynthfinder.search.retrostar.nodes.ReactionNode(cost, reaction, parent)

Bases: aizynthfinder.search.andor_trees.TreeNodeMixin

An AND node representing a reaction

Variables

  • cost – the cost of the reaction

  • parent – the parent of the node

  • reaction – the reaction represented by the node

  • solved – if True all children nodes are solved

  • target_value – the V(m|T) for the children, the current cost

  • value – the current rn(r|T)

Parameters

  • cost (float) – the cost of the reaction

  • reaction (RetroReaction) – the reaction to be represented by the node

  • parent (MoleculeNode) – the parent of the node

Return type

None

classmethod create_stub(cost, reaction, parent, config)

Create a ReactionNode and creates all the MoleculeNode objects that are the children of the node.

Parameters

  • cost (float) – the cost of the reaction

  • reaction (RetroReaction) – the reaction to be represented by the node

  • parent (MoleculeNode) – the parent of the node

  • config (Configuration) – the configuration of the search tree

Return type

ReactionNode

classmethod from_dict(dict_, config, molecules, molecule_cost, parent)

Create a new node from a dictionary, i.e. deserialization

Parameters
Returns

a deserialized node

Return type

ReactionNode

property children: List[MoleculeNode]

Gives the molecule children nodes

property prop: StrDict

Dictionary with publicly exposed properties

serialize(molecule_store)

Serialize the node object to a dictionary

Parameters

molecule_store (MoleculeSerializer) – the serialized molecules

Returns

the serialized node

Return type

StrDict

update(value, from_mol=None)

Update the node as part of the update algorithm, calling the update() method of its parent

Parameters

  • value (float) – the delta V value

  • from_mol (Optional[TreeMolecule]) – the molecule being expanded, used for excluding propagation

Return type

None

aizynthfinder.search.retrostar.search_tree module

Module containing a class that holds the tree search

class aizynthfinder.search.retrostar.search_tree.SearchTree(config, root_smiles=None)

Bases: aizynthfinder.search.andor_trees.AndOrSearchTreeBase

Encapsulation of the Retro* search tree (an AND/OR tree).

Variables

  • config – settings of the tree search algorithm

  • root – the root node

Parameters

  • config (Configuration) – settings of the tree search algorithm

  • root_smiles (Optional[str]) – the root will be set to a node representing this molecule, defaults to None

Return type

None

classmethod from_json(filename, config)

Create a new search tree by deserialization from a JSON file

Parameters

  • filename (str) – the path to the JSON node

  • config (Configuration) – the configuration of the search tree

Returns

a deserialized tree

Return type

SearchTree

property mol_nodes: Sequence[MoleculeNode]

Return the molecule nodes of the tree

one_iteration()
Perform one iteration of

  1. Selection

  2. Expansion

  3. Update

Raises

StopIteration – if the search should be pre-maturely terminated

Returns

if a solution was found

Return type

bool

routes()

Extracts and returns routes from the AND/OR tree

Returns

the routes

Return type

List[ReactionTree]

serialize(filename)

Seralize the search tree to a JSON file

Parameters

filename (str) – the path to the JSON file

Return type

None

Module contents