Source code for schrodinger.livedesign.convert

import enum
from functools import partial
from typing import Generator
from typing import Union

from rdkit import Chem
from rdkit.Chem import rdChemReactions
from rdkit.Chem import rdmolops

from schrodinger.thirdparty import rdkit_adapter

ATOM_PROP_ATOM_LABEL = "atomLabel"
BOND_PROP_BOND_STEREO = "_MolFileBondStereo"
BOND_PROP_BOND_CONFIG = "_MolFileBondCfg"
MOL_PROP_ATTACHPT = "molAttchpt"
MOL_PROP_R_LABEL = "_MolFileRLabel"


[docs]class Format(enum.Enum): SDF = enum.auto() SMILES = enum.auto() CXSMILES = enum.auto() SMARTS = enum.auto() INCHI = enum.auto()
[docs]def convert(data: str, input_format: Format, output_format: Format) -> str: """ Main entrypoint for converting one serialization to another. A few convience functions are provided below for common LD conversions. :param data: input text string :param input_format: expected format for input string :param output_format: desired format for output string :return: converted text string """ if _is_reaction(data, input_format): return _from_rdkit_reaction(_to_rdkit_reaction(data, format=input_format), format=output_format) else: return _from_rdkit_mol(_to_rdkit_mol(data, format=input_format), format=output_format)
[docs]def sdf_to_rdkit( molblock: str ) -> Union[Chem.rdchem.Mol, Chem.rdChemReactions.ChemicalReaction]: """ :param molblock: given SDF or RXN molblock :return: corresponding RDKit mol or reaction """ if _is_reaction(molblock, Format.SDF): return _to_rdkit_reaction(molblock, format=Format.SDF) else: return _to_rdkit_mol(molblock, format=Format.SDF)
[docs]def rdkit_to_sdf(mol: Chem.rdchem.Mol, kekulize: bool = True) -> str: """ :param input: given RDKit mol :param kekulize: whether to kekulize mol :return: corresponding SDF molblock """ return _from_rdkit_mol(mol, format=Format.SDF, kekulize_sdf=kekulize)
[docs]def rdkit_to_cxsmiles(mol: Chem.rdchem.Mol) -> str: """ :param mol: given RDKit mol :return: corresponding CXSMILES stripped of coordinates """ return _from_rdkit_mol(mol, format=Format.CXSMILES)
[docs]def get_sd_reader(molblock: str) -> Generator[Chem.rdchem.Mol, None, None]: """ :param molblock: string of SDF molblocks :return: generator that provides mols from SDF molblcoks """ reader = Chem.SDMolSupplier() reader.SetData(molblock, sanitize=False, removeHs=False, strictParsing=False) for mol in reader: yield _replace_attachment_points(mol)
[docs]def atom_has_attachment_point(atom): return atom.HasProp(ATOM_PROP_ATOM_LABEL) and atom.GetProp( ATOM_PROP_ATOM_LABEL).startswith("_AP")
def _is_reaction(data: str, format: Format) -> bool: if format == Format.SDF: return data.startswith("$RXN") elif format in (Format.SMILES, Format.CXSMILES, Format.SMARTS): return ">>" in data else: return False # Unsupported reaction format def _to_rdkit_mol(data: str, format: Format) -> Chem.rdchem.Mol: def _from_sdf(molblock: str) -> Chem.rdchem.Mol: reader = get_sd_reader(molblock) mol = next(reader) if mol is None: raise RuntimeError("Single molblock required; 0 present") for m in reader: # there is another molblock if this can loop raise RuntimeError("Single molblock required; multiple present") return mol _from_smiles = partial(Chem.MolFromSmiles, sanitize=False) str_to_mol = { Format.SDF: _from_sdf, Format.SMILES: _from_smiles, Format.CXSMILES: _from_smiles, Format.SMARTS: Chem.MolFromSmarts, Format.INCHI: partial(Chem.MolFromInchi, sanitize=False), } with rdkit_adapter.convert_log_to_exception(): mol = str_to_mol[format](data) _fix_r0_rgroup(mol) # partial sanitization to preserve input molecule ops = (Chem.SANITIZE_ALL ^ Chem.SANITIZE_CLEANUP ^ Chem.SANITIZE_PROPERTIES ^ Chem.SANITIZE_KEKULIZE ^ Chem.SANITIZE_FINDRADICALS ^ Chem.SANITIZE_CLEANUPCHIRALITY) Chem.SanitizeMol(mol, sanitizeOps=ops) mol.UpdatePropertyCache(strict=False) return mol def _from_rdkit_mol(mol: Chem.rdchem.Mol, format: Format, kekulize_sdf: bool = True) -> str: def _to_cxsmiles(mol: Chem.rdchem.Mol) -> str: mol.RemoveAllConformers() # remove coordinates from CXSMILES return Chem.MolToCXSmiles(mol) def _to_sdf(mol: Chem.rdchem.Mol, kekulize: bool) -> str: mol = _replace_dummy_atoms_with_attchpts(mol) # SHARED-8282, SHARED-8498: there's cases that can't be kekulized, # so just kekulize if possible if kekulize: Chem.rdmolops.KekulizeIfPossible(mol, clearAromaticFlags=True) return Chem.SDWriter.GetText(mol, kekulize=False, force_v3000=True) mol_to_str = { Format.SDF: partial(_to_sdf, kekulize=kekulize_sdf), Format.SMILES: Chem.MolToSmiles, Format.SMARTS: Chem.MolToSmarts, Format.CXSMILES: _to_cxsmiles, Format.INCHI: Chem.MolToInchi, } return mol_to_str[format](Chem.Mol(mol)) def _to_rdkit_reaction(data: str, format: Format) -> Chem.rdChemReactions.ChemicalReaction: str_to_rxn = { Format.SDF: rdChemReactions.ReactionFromRxnBlock, Format.SMILES: partial(rdChemReactions.ReactionFromSmarts, useSmiles=True), Format.SMARTS: rdChemReactions.ReactionFromSmarts, Format.CXSMILES: partial(rdChemReactions.ReactionFromSmarts, useSmiles=True), } # RDKit doesn't support SMARTS extensions in reactions if format in (Format.SMARTS, Format.SMILES, Format.CXSMILES): data = data.split(" ")[0] if format == Format.INCHI: raise RuntimeError("Cannot convert INCHI format to reaction") rxn = str_to_rxn[format](data) for reactant in rxn.GetReactants(): _fix_r0_rgroup(reactant) for product in rxn.GetProducts(): _fix_r0_rgroup(product) return rxn def _from_rdkit_reaction(rxn: Chem.rdChemReactions.ChemicalReaction, format: Format) -> str: rxn_to_str = { Format.SDF: rdChemReactions.ReactionToRxnBlock, Format.SMILES: rdChemReactions.ReactionToSmiles, Format.SMARTS: rdChemReactions.ReactionToSmarts, Format.CXSMILES: rdChemReactions.ReactionToSmiles, } if format == Format.INCHI: raise RuntimeError("Cannot convert reaction to INCHI format") return rxn_to_str[format](rxn) def _set_attachment_point_coordinates(mol, edit_mol, attch_pts): """ Calculates coordinate of recently added attachment points by using coordinates of implicit hydrogens in the mol and assigns chiral tags using bond directions and new coordinates """ mol.UpdatePropertyCache(strict=False) temp_mol = Chem.AddHs(mol, addCoords=True, onlyOnAtoms=attch_pts.keys()) temp_mol_positions = temp_mol.GetConformer().GetPositions() for parent, attch_pt_list in attch_pts.items(): parent_atom = temp_mol.GetAtomWithIdx(parent) added_hydrogens = (h for h in parent_atom.GetNeighbors() if h.GetIdx() >= mol.GetNumAtoms()) for at_idx, h in zip(attch_pt_list, added_hydrogens): # h is a graph hydrogen whose coordinates we can use for # the recently added attachment point xyz = temp_mol_positions[h.GetIdx()] edit_mol.GetConformer().SetAtomPosition(at_idx, xyz) mol = reapply_molblock_wedging(edit_mol) Chem.AssignChiralTypesFromBondDirs(mol) return Chem.Mol(mol) def _replace_attachment_points(mol): """ Replace attachment points with dummy atoms to preserve chirality """ if mol is None: return None attch_pts = { a.GetIdx(): [] for a in mol.GetAtoms() if a.HasProp(MOL_PROP_ATTACHPT) } if not attch_pts: return mol # Replace attachment points with dummy atoms edit_mol = Chem.RWMol(mol) attch_pt_num = 1 for atom_idx in attch_pts: num_attch_pts = 1 parent_atom = edit_mol.GetAtomWithIdx(atom_idx) attch_num = parent_atom.GetIntProp(MOL_PROP_ATTACHPT) if attch_num == -1: # this atom has two attachment points num_attch_pts = 2 attch_num = 1 exp_hs = parent_atom.GetNumExplicitHs() for i in range(num_attch_pts): new_at = edit_mol.AddAtom(Chem.Atom(0)) attch_pts[atom_idx].append(new_at) edit_mol.AddBond(atom_idx, new_at, Chem.BondType.SINGLE) edit_mol.GetAtomWithIdx(new_at).SetProp(ATOM_PROP_ATOM_LABEL, f'_AP{attch_num}') attch_num += 1 parent_atom.ClearProp(MOL_PROP_ATTACHPT) if exp_hs: parent_atom.SetNumExplicitHs(exp_hs - num_attch_pts) # Dummy atoms should be query atoms. This ensures the dummy atom is written # as '*' instead of 'R' params = rdmolops.AdjustQueryParameters.NoAdjustments() params.makeDummiesQueries = True edit_mol = rdmolops.AdjustQueryProperties(edit_mol, params) return _set_attachment_point_coordinates(mol, edit_mol, attch_pts) def _replace_dummy_atoms_with_attchpts(mol): # According to the MDL specification, ATTCHPT=1 is the first attachment point, # ATTCHPT=2 is the second attachment point, and ATTCHPT=-1 means that this # atom has the first and second attachment points. remove_attchpts = [] for at in mol.GetAtoms(): if atom_has_attachment_point(at): remove_attchpts.append(at.GetIdx()) parent = at.GetNeighbors()[0] exp_hs = parent.GetNumExplicitHs() if parent.HasProp(MOL_PROP_ATTACHPT): parent.SetIntProp(MOL_PROP_ATTACHPT, -1) parent.SetNumExplicitHs(exp_hs + 2) else: parent.SetIntProp(MOL_PROP_ATTACHPT, int(at.GetProp(ATOM_PROP_ATOM_LABEL)[3])) parent.SetNumExplicitHs(exp_hs + 1) if not remove_attchpts: return mol # remove attachment points edit_mol = Chem.RWMol(mol) edit_mol.BeginBatchEdit() for at_idx in remove_attchpts: edit_mol.RemoveAtom(at_idx) edit_mol.CommitBatchEdit() return edit_mol.GetMol() def _detect_unkekulized_atoms(m): opts = Chem.SANITIZE_ALL ^ Chem.SANITIZE_PROPERTIES ^ Chem.SANITIZE_CLEANUP ^ Chem.SANITIZE_CLEANUPCHIRALITY ^ Chem.SANITIZE_FINDRADICALS errs = Chem.DetectChemistryProblems(m, opts) for err in errs: if err.GetType() == 'KekulizeException': return sorted(err.GetAtomIndices()) return None
[docs]def reapply_molblock_wedging(mol): for bnd in mol.GetBonds(): # only change the wedging if the bond was wedged in the input data (we # recognize that by looking for the properties the mol file parser # sets): if bnd.HasProp(BOND_PROP_BOND_STEREO): # V2000 val = bnd.GetProp(BOND_PROP_BOND_STEREO) if val == "1": bnd.SetBondDir(Chem.BondDir.BEGINWEDGE) elif val == "6": bnd.SetBondDir(Chem.BondDir.BEGINDASH) elif val == "4": bnd.SetBondDir(Chem.BondDir.UNKNOWN) elif bnd.HasProp(BOND_PROP_BOND_CONFIG): # V3000 val = bnd.GetProp(BOND_PROP_BOND_CONFIG) if val == "1": bnd.SetBondDir(Chem.BondDir.BEGINWEDGE) elif val == "3": bnd.SetBondDir(Chem.BondDir.BEGINDASH) elif val == "2": bnd.SetBondDir(Chem.BondDir.UNKNOWN) elif bnd.GetBondDir() in (Chem.BondDir.BEGINDASH, Chem.BondDir.BEGINWEDGE, Chem.BondDir.UNKNOWN): bnd.SetBondDir(Chem.BondDir.NONE) return mol
[docs]def add_hs_to_aromatic_nitrogen(mol): """ Intended to be used with molecules which have kekulization failures due to aromatic system(s) containing Ns where the user hasn't provided the location of the implicit Hs in the system. This picks an arbitrary (but canonical) aromatic N to add an H to in each aromatic system. Returns the original mol if it can't fix it. """ Chem.GetSymmSSSR(mol) tm = Chem.Mol(mol) unkekulized_atoms = _detect_unkekulized_atoms(tm) while unkekulized_atoms: tm, new_unkekulized_atoms = _fix_ring_system(tm, unkekulized_atoms) if new_unkekulized_atoms == unkekulized_atoms: # got stuck, return the original return mol unkekulized_atoms = new_unkekulized_atoms return tm
def _fix_ring_system(m, unkekulized_atoms): m.UpdatePropertyCache(False) # we want to always generate the same result for the same molecule # Which result to return will cause arguments, yay, but we'll # prefer smaller rings, then use the canonical atom ranking to break ties ranks = Chem.CanonicalRankAtoms(m) ri = m.GetRingInfo() atoms_to_check = unkekulized_atoms[:] atoms_to_check.sort(key=lambda x: (ri.MinAtomRingSize(x), ranks[x])) while atoms_to_check: ai = atoms_to_check.pop(0) atom = m.GetAtomWithIdx(ai) if atom.GetAtomicNum() != 7: continue if atom.GetDegree() != 2 or atom.GetNumExplicitHs() != 0: continue tm = Chem.Mol(m) atom = tm.GetAtomWithIdx(ai) atom.SetNumExplicitHs(1) atom.SetNoImplicit(True) # did we fix something? new_unkekulized_atoms = _detect_unkekulized_atoms(tm) if new_unkekulized_atoms != unkekulized_atoms: return tm, new_unkekulized_atoms return m, unkekulized_atoms def _fix_r0_rgroup(mol): highest_rlabel = 0 atoms_with_r0 = [] for atom in mol.GetAtoms(): try: label = atom.GetIntProp(MOL_PROP_R_LABEL) except KeyError: continue highest_rlabel = max(highest_rlabel, label) if label == 0: atoms_with_r0.append(atom) if atoms_with_r0: highest_rlabel += 1 for atom in atoms_with_r0: atom.SetIntProp(MOL_PROP_R_LABEL, highest_rlabel)