xenosite-fragment 0a3

Library for processing molecule fragments.

Create a fragment from a SMILES or a fragment SMILES string:

str(Fragment("CCCC")) # Valid smiles 'C-C-C-C'

str(Fragment("ccCC")) # not a valid SMILES 'c:c-C-C'

Optionally, create a fragment of a molecule from a string and (optionally) a list of nodes in the fragment.

F = Fragment("CCCCCCOc1ccccc1", [0,1,2,3,4,5]) str(F) # hexane 'C-C-C-C-C-C'

If IDs are provided, they MUST select a connected fragment.

F = Fragment("CCCCCCOc1ccccc1", [0,10]) Traceback (most recent call last): ... ValueError: Multiple components in graph are not allowed.

Get the canonical representation of a fragment:

Fragment("O-C").canonical().string 'C-O' Fragment("OC").canonical().string 'C-O' Fragment("CO").canonical().string 'C-O'

Get the reordering of nodes used to create the canonical string representation. If remap=True, then the ID are remapped to the input representation used to initalize the Fragment.

Fragment("COC", [1,2]).canonical(remap=True).reordering [2, 1] Fragment("COC", [1,2]).canonical().reordering [1, 0]

Match fragment to a molecule. By default, the ID correspond with fragment IDs. If remap=True, the ID corresponds to the input representation when the Fragment was initialized.

smiles = "CCCC1CCOCN1" F = Fragment("CCCCCC") # hexane as a string list(F.matches(smiles)) # smiles string (least efficient) [(0, 1, 2, 3, 4, 5)]

import rdkit mol = rdkit.Chem.MolFromSmiles(smiles) list(F.matches(mol)) # RDKit mol [(0, 1, 2, 3, 4, 5)]

mol_graph = Graph.from_molecule(mol) list(F.matches(mol, mol_graph)) # RDKit mol and Graph (most efficient) [(0, 1, 2, 3, 4, 5)]

Matches ensure that the fragment string of matches is the same as the fragment. This is different than standards SMARTS matching, and prevents rings from matching unclosed ring patterns:

str(Fragment("C1CCCCC1")) # cyclohexane 'C1-C-C-C-C-C-1'

assert(str(Fragment("C1CCCCC1")) != str(F)) # cyclohexane is not hexane list(F.matches("C1CCCCC1")) # Unlike SMARTS, no match! []

Efficiently create multiple fragments by reusing a precomputed graph:

import rdkit

mol = rdkit.Chem.MolFromSmiles("c1ccccc1OCCC") mol_graph = Graph.from_molecule(mol)

f1 = Fragment(mol_graph, [0]) f2 = Fragment(mol_graph, [6,5,4])

Find matches to fragments:

list(f1.matches(mol)) [(0,), (1,), (2,), (3,), (4,), (5,)]

list(f2.matches(mol)) [(6, 5, 4), (6, 5, 0)]

Fragments know how to report if they are canonically the same as each other or strings.

Fragment("CCO") == Fragment("OCC") True Fragment("CCO") == "C-C-O" True

Note, however, that strings are not converted to canonical form. Therefore,

Fragment("CCO") == "CCO" False

Enumerate and compute statistics on all the subgraphs in a molecule:

from xenosite.fragment.net import SubGraphFragmentNetwork N = SubGraphFragmentNetwork("CC1COC1") fragments = N.to_pandas() list(fragments.index) ['C-C', 'C', 'C-O-C', 'C-O', 'O', 'C-C1-C-O-C-1', 'C1-C-O-C-1', 'C-C-C-O', 'C-C(-C)-C', 'C-C-O', 'C-C-C'] fragments["size"].to_numpy() array([2, 1, 3, 2, 1, 5, 4, 4, 4, 3, 3])

Better fragments can be enumerated by collapsing all atoms in a ring into a single node during subgraph enumeration.

from xenosite.fragment.net import RingFragmentNetwork N = RingFragmentNetwork("CC1COC1") fragments = N.to_pandas() list(fragments.index) ['C-C1-C-O-C-1', 'C', 'C1-C-O-C-1'] fragments["size"].to_numpy() array([5, 1, 4])