1.1.1.2.1.1. molpher.algorithms.antidecoys package

1.1.1.2.1.1.1. Submodules

1.1.1.2.1.1.2. molpher.algorithms.antidecoys.custom_opers module

Custom operations and callbacks used in AntidecoysPathFinder.

class molpher.algorithms.antidecoys.custom_opers.AntiFpSortCallback(antifp_scores)[source]

Bases: molpher.swig_wrappers.core.SortMorphsCallback

Parameters

antifp_scores (dict) – SMILES to anti-fingerprint similarity map (for each molecule in candidates)

This callback is used to sort candidates with regard to their similarity to the anti-fingerprint. Molecules less similar to the anti-fingerprint are ranked higher than more similar ones.

antifp_scores

SMILES to anti-fingerprint score map (for each molecule in candidates)

maximum_common_bits_perc

maximum percentage of bits that one candidate morph has in common with the anti-fingerprint

minimum_common_bits_perc

minimum percentage of bits that one candidate morph has in common with the anti-fingerprint

class molpher.algorithms.antidecoys.custom_opers.AntidecoysFilter(scores, common_bits_thrs, minimum_accept)[source]

Bases: molpher.core.operations.TreeOperation.TreeOperation

Parameters

  • scores – a dict mapping a SMILES string of a molecule in candidates to the corresponding score value

  • common_bits_thrs – maximum percentage of bits in common with the anti-fingerprint allowed in the

  • minimum_accept – always accept at least this many top scoring molecules

A filtering tree operation which filters the morphs in candidates based on common_bits_thrs. All morphs with anti-fingerprint score bigger than common_bits_thrs are rejected. However, at least minimum_accept best morphs survives.

common_bits_thrs

maximum percentage of bits in common with the anti-fingerprint allowed in the

minimum_accept

always accept at least this many top scoring molecules

scores

a dict mapping a SMILES string of a molecule in candidates to the corresponding score value

class molpher.algorithms.antidecoys.custom_opers.GatherAntiFPScores(out, anti_fp, settings)[source]

Bases: molpher.core.operations.TreeOperation.TreeOperation

Parameters

  • out – a dictionary to save the results to

  • anti_fp – the anti-fingerprint

  • settings – AntidecoysSettings

A tree operation which computes the anti-fingerprint scores using the given settings and anti-fingerprint. For every molecule in candidates the similarity to the anti-fingerprint is computed and saved as a score into the supplied dictionary.

See also

compute_antifp_scores

antitfp

the anti-fingerprint

out

output dictionary

settings

AntidecoysSettings

1.1.1.2.1.1.3. molpher.algorithms.antidecoys.pathfinder module

class molpher.algorithms.antidecoys.pathfinder.AntidecoysPathFinder(settings, antifingerprint=None)[source]

Bases: object

Parameters

  • settings (AntidecoysSettings) – exploration settings and parameters

  • antifingerprint (rdkit.cDataStructs.SparseBitVector) – the anti-fingerprint to use (optional)

Implements a modified version of the bidirectional algorithm. The algorithm tries to avoid some already sampled areas of chemical space by using an ‘anti-fingerprint’, a 2D pharmacophore fingerprint that describes previously explored pharmacophore features and their geometric relationships.

The RDKit implementation of pharmacophore fingerprints is used. Therefore, it is necessary to have RDKit installed to use this class and this module.

If no antifingerprint is specified, the search defaults to an ordinary bidirectional algorithm.

__call__()[source]

Execute the search

Returns

list of SMILES strings representing the path found

__dict__ = mappingproxy({'__module__': 'molpher.algorithms.antidecoys.pathfinder', '__doc__': "\n    :param settings: exploration settings and parameters\n    :type settings: `AntidecoysSettings`\n    :param antifingerprint: the anti-fingerprint to use (optional)\n    :type antifingerprint: :class:`rdkit.cDataStructs.SparseBitVector`\n\n    Implements a modified version of the bidirectional algorithm. The algorithm tries to avoid\n    some already sampled areas of chemical space by using an 'anti-fingerprint',\n    a 2D pharmacophore fingerprint that describes previously explored pharmacophore features and their geometric relationships.\n\n    The `RDKit <http://www.rdkit.org/docs/RDKit_Book.html#representation-of-pharmacophore-fingerprints>`_\n    implementation of pharmacophore fingerprints is used. Therefore, it is necessary to have RDKit installed\n    to use this class and this module.\n\n    If no :samp:`antifingerprint` is specified, the search defaults to an ordinary :mod:`~molpher.algorithms.bidirectional` algorithm.\n\n    ", '__init__': <function AntidecoysPathFinder.__init__>, '__call__': <function AntidecoysPathFinder.__call__>, '__dict__': <attribute '__dict__' of 'AntidecoysPathFinder' objects>, '__weakref__': <attribute '__weakref__' of 'AntidecoysPathFinder' objects>, '__annotations__': {}})
__init__(settings, antifingerprint=None)[source]

Initialize self. See help(type(self)) for accurate signature.

__module__ = 'molpher.algorithms.antidecoys.pathfinder'
__weakref__

list of weak references to the object (if defined)

antifingerprint

the anti-fingerprint

connecting_molecule

SMILES string of a molecule that connects the two trees

path

a list of SMILES strings representing the found path (defaults to an empty list)

settings

AntidecoysSettings object used to initialize this instance

source

SMILES of the source

source_target

tree searching from source to target

source_target_min

FindClosest holding the current minimum in the ‘source to target’ tree

target

SMILES of the target

target_source

tree searching from target to source

target_source_min

FindClosest holding the current minimum in the ‘target to source’ tree

verbose

use verbose output

1.1.1.2.1.1.4. molpher.algorithms.antidecoys.run module

molpher.algorithms.antidecoys.run.run(settings, paths_count)[source]
Parameters

  • settings – instance of AntidecoysSettings specifying search parameters

  • paths_count – number of paths to generate

Returns

a list of lists of SMILES strings (each path is represented by a list of SMILES strings)

1.1.1.2.1.1.5. molpher.algorithms.antidecoys.settings module

class molpher.algorithms.antidecoys.settings.AntidecoysSettings(source, target, storage_dir='/home/sichom/projects/molpher-lib/doc/antidecoys_data', max_threads=None, tree_params=None, max_iters=100, verbose=False, fg_bins=((0, 2), (2, 5), (5, 8)), fg_min_points=2, fg_max_points=3, min_accepted=1000, common_bits_max_thrs=0.75, common_bits_mean_thrs=0.5, antidecoys_min_iters=10, antidecoys_max_iters=50, distance_thrs=0.2)[source]

Bases: molpher.algorithms.settings.Settings

A specialized class that holds the settings for the anti-decoys algorithm.

antidecoys_max_iters

maximum number of iterations that will use the antidecoys filter

antidecoys_min_iters

minimum number of iterations that will use the antidecoys filter

common_bits_max_thrs

maximum percentage of shared bits between a structure and the anti-fingerprint that the filter will accept on every iteration

common_bits_mean_thrs

if for the structures that survived the filter the mean common bits percentage falls below this value, antidecoys will be turned off

distance_thrs

turn antidecoys filter off when the distance between two closest molecules from each tree gets below this value

fg_bins

distance bins in the pharmacophore fingerprint (as described here)

fg_max_points

max number of features encoded in the pharmacophore fingerprint (as described here)

fg_min_points

min number of features encoded in the pharmacophore fingerprint (as described here)

min_accepted

minimum number of morphs the antidecoys filter will accept on every iteration

1.1.1.2.1.1.6. molpher.algorithms.antidecoys.utils module

molpher.algorithms.antidecoys.utils.compute_anti_fp(mols_smiles, sig_fac, antifp_old=None)[source]

Computes an anti-fingerprint from the given molecules.

It is possible to specify an existing anti-fingerprint for an update. In this case, the returned fingerprint will be the result of doing a bitwise or between the old fingerprint and the one generated from the supplied structures.

Parameters

  • mols_smiles – SMILES of the molecules to generate the anti-fingerprint from

  • sig_fac – RDKit’s signature factory used in the 2D pharmacophore fingerprint computation

  • antifp_old – an old anti-fingerprint to update

Returns

new or updated anti-fingerprint

molpher.algorithms.antidecoys.utils.compute_antifp_scores(mols, anti_fp, settings)[source]

Computes (in parallel) the anti-fingerprint scores for the given molecules.

Parameters

  • mols – molecules as a list of SMILES strings

  • anti_fp – the anti-fingerprint

  • settings – AntidecoysSettings

Returns

a dict mapping SMILES of the given molecules to the anti-fingerprint scores

molpher.algorithms.antidecoys.utils.update_target(tree, target)[source]

Change the tree’s target molecule.

Parameters

1.1.1.2.1.1.7. Module contents