// find the neighbors for an atoms that are not connected by single bond that is
// not refBond
// if checkDir is true only neighbor atoms with bonds marked with a direction
// will be returned
void findAtomNeighborsHelper(const ROMol &mol, const Atom *atom,
const Bond *refBond, UINT_VECT &neighbors,
bool checkDir = false) {
PRECONDITION(atom, "bad atom");
PRECONDITION(refBond, "bad bond");
neighbors.clear();
ROMol::OEDGE_ITER beg, end;
boost::tie(beg, end) = mol.getAtomBonds(atom);
while (beg != end) {
const BOND_SPTR bond = mol[*beg];
Bond::BondDir dir = bond->getBondDir();
if (bond->getBondType() == Bond::SINGLE &&
bond->getIdx() != refBond->getIdx()) {
if (checkDir) {
if ((dir != Bond::ENDDOWNRIGHT) && (dir != Bond::ENDUPRIGHT)) {
++beg;
continue;
}
}
Atom *nbrAtom = bond->getOtherAtom(atom);
neighbors.push_back(nbrAtom->getIdx());
}
++beg;
}
}
std::string MolToSmarts(ROMol &inmol, bool doIsomericSmiles) {
RDUNUSED_PARAM(doIsomericSmiles); // does this parameter even make sense?
std::string res;
unsigned int nAtoms = inmol.getNumAtoms();
if (!nAtoms) return "";
ROMol mol(inmol);
UINT_VECT ranks;
ranks.resize(nAtoms);
// For smiles writing we would be canonicalizing but we will not do that here.
// We will simple use the atom indices as the rank
for (ROMol::AtomIterator atIt = mol.beginAtoms(); atIt != mol.endAtoms();
atIt++) {
ranks.push_back((*atIt)->getIdx());
}
std::vector<AtomColors> colors;
colors.resize(nAtoms);
std::vector<AtomColors>::iterator colorIt;
for (colorIt = colors.begin(); colorIt != colors.end(); colorIt++)
*colorIt = Canon::WHITE_NODE;
colorIt = std::find(colors.begin(), colors.end(), Canon::WHITE_NODE);
while (colorIt != colors.end()) {
unsigned int nextAtomIdx = 0;
unsigned int nextRank;
std::string subSmi;
nextRank = nAtoms + 1;
for (unsigned int i = 0; i < nAtoms; i++) {
if (colors[i] == Canon::WHITE_NODE && ranks[i] < nextRank) {
nextRank = ranks[i];
nextAtomIdx = i;
}
}
subSmi = FragmentSmartsConstruct(mol, nextAtomIdx, colors, ranks);
res += subSmi;
colorIt = std::find(colors.begin(), colors.end(), Canon::WHITE_NODE);
if (colorIt != colors.end()) {
res += ".";
}
}
return res;
}
// Figure out the CIP ranks for the atoms of a molecule
void assignAtomCIPRanks(const ROMol &mol, UINT_VECT &ranks) {
PRECONDITION((!ranks.size() || ranks.size() >= mol.getNumAtoms()),
"bad ranks size");
if (!ranks.size()) ranks.resize(mol.getNumAtoms());
unsigned int numAtoms = mol.getNumAtoms();
#ifndef USE_NEW_STEREOCHEMISTRY
// get the initial invariants:
DOUBLE_VECT invars(numAtoms, 0);
buildCIPInvariants(mol, invars);
iterateCIPRanks(mol, invars, ranks, false);
#else
Canon::chiralRankMolAtoms(mol, ranks);
#endif
// copy the ranks onto the atoms:
for (unsigned int i = 0; i < numAtoms; ++i) {
mol[i]->setProp(common_properties::_CIPRank, ranks[i], 1);
}
}
void iterateCIPRanks(const ROMol &mol, DOUBLE_VECT &invars, UINT_VECT &ranks,
bool seedWithInvars) {
PRECONDITION(invars.size() == mol.getNumAtoms(), "bad invars size");
PRECONDITION(ranks.size() >= mol.getNumAtoms(), "bad ranks size");
unsigned int numAtoms = mol.getNumAtoms();
CIP_ENTRY_VECT cipEntries(numAtoms);
INT_LIST allIndices;
for (unsigned int i = 0; i < numAtoms; ++i) {
allIndices.push_back(i);
}
#ifdef VERBOSE_CANON
BOOST_LOG(rdDebugLog) << "invariants:" << std::endl;
for (unsigned int i = 0; i < numAtoms; i++) {
BOOST_LOG(rdDebugLog) << i << ": " << invars[i] << std::endl;
}
#endif
// rank those:
Rankers::rankVect(invars, ranks);
#ifdef VERBOSE_CANON
BOOST_LOG(rdDebugLog) << "initial ranks:" << std::endl;
for (unsigned int i = 0; i < numAtoms; ++i) {
BOOST_LOG(rdDebugLog) << i << ": " << ranks[i] << std::endl;
}
#endif
// Start each atom's rank vector with its atomic number:
// Note: in general one should avoid the temptation to
// use invariants here, those lead to incorrect answers
for (unsigned int i = 0; i < numAtoms; i++) {
if (!seedWithInvars) {
cipEntries[i].push_back(mol[i]->getAtomicNum());
cipEntries[i].push_back(static_cast<int>(ranks[i]));
} else {
cipEntries[i].push_back(static_cast<int>(invars[i]));
}
}
// Loop until either:
// 1) all classes are uniquified
// 2) the number of ranks doesn't change from one iteration to
// the next
// 3) we've gone through maxIts times
// maxIts is calculated by dividing the number of atoms
// by 2. That's a pessimal version of the
// maximum number of steps required for two atoms to
// "feel" each other (each influences one additional
// neighbor shell per iteration).
unsigned int maxIts = numAtoms / 2 + 1;
unsigned int numIts = 0;
int lastNumRanks = -1;
unsigned int numRanks = *std::max_element(ranks.begin(), ranks.end()) + 1;
while (numRanks < numAtoms && numIts < maxIts &&
(lastNumRanks < 0 ||
static_cast<unsigned int>(lastNumRanks) < numRanks)) {
unsigned int longestEntry = 0;
// ----------------------------------------------------
//
// for each atom, get a sorted list of its neighbors' ranks:
//
for (INT_LIST_I it = allIndices.begin(); it != allIndices.end(); ++it) {
CIP_ENTRY localEntry;
localEntry.reserve(16);
// start by pushing on our neighbors' ranks:
ROMol::OEDGE_ITER beg, end;
boost::tie(beg, end) = mol.getAtomBonds(mol[*it].get());
while (beg != end) {
const Bond *bond = mol[*beg].get();
++beg;
unsigned int nbrIdx = bond->getOtherAtomIdx(*it);
const Atom *nbr = mol[nbrIdx].get();
int rank = ranks[nbrIdx] + 1;
// put the neighbor in 2N times where N is the bond order as a double.
// this is to treat aromatic linkages on fair footing. i.e. at least in
// the
// first iteration --c(:c):c and --C(=C)-C should look the same.
// this was part of issue 3009911
unsigned int count;
if (bond->getBondType() == Bond::DOUBLE && nbr->getAtomicNum() == 15 &&
(nbr->getDegree() == 4 || nbr->getDegree() == 3)) {
// a special case for chiral phophorous compounds
// (this was leading to incorrect assignment of
// R/S labels ):
count = 1;
// general justification of this is:
// Paragraph 2.2. in the 1966 article is "Valence-Bond Conventions:
// Multiple-Bond Unsaturation and Aromaticity". It contains several
// conventions of which convention (b) is the one applying here:
// "(b) Contibutions by d orbitals to bonds of quadriligant atoms are
// neglected."
// FIX: this applies to more than just P
} else {
count = static_cast<unsigned int>(
floor(2. * bond->getBondTypeAsDouble() + .1));
}
CIP_ENTRY::iterator ePos =
std::lower_bound(localEntry.begin(), localEntry.end(), rank);
localEntry.insert(ePos, count, rank);
++nbr;
}
// add a zero for each coordinated H:
// (as long as we're not a query atom)
if (!mol[*it]->hasQuery()) {
localEntry.insert(localEntry.begin(), mol[*it]->getTotalNumHs(), 0);
}
// we now have a sorted list of our neighbors' ranks,
// copy it on in reversed order:
cipEntries[*it].insert(cipEntries[*it].end(), localEntry.rbegin(),
localEntry.rend());
if (cipEntries[*it].size() > longestEntry) {
longestEntry = rdcast<unsigned int>(cipEntries[*it].size());
}
}
// ----------------------------------------------------
//
// pad the entries so that we compare rounds to themselves:
//
for (INT_LIST_I it = allIndices.begin(); it != allIndices.end(); ++it) {
unsigned int sz = rdcast<unsigned int>(cipEntries[*it].size());
if (sz < longestEntry) {
cipEntries[*it].insert(cipEntries[*it].end(), longestEntry - sz, -1);
}
}
// ----------------------------------------------------
//
// sort the new ranks and update the list of active indices:
//
lastNumRanks = numRanks;
Rankers::rankVect(cipEntries, ranks);
numRanks = *std::max_element(ranks.begin(), ranks.end()) + 1;
// now truncate each vector and stick the rank at the end
for (unsigned int i = 0; i < numAtoms; ++i) {
cipEntries[i][numIts + 1] = ranks[i];
cipEntries[i].erase(cipEntries[i].begin() + numIts + 2,
cipEntries[i].end());
}
++numIts;
#ifdef VERBOSE_CANON
BOOST_LOG(rdDebugLog) << "strings and ranks:" << std::endl;
for (unsigned int i = 0; i < numAtoms; i++) {
BOOST_LOG(rdDebugLog) << i << ": " << ranks[i] << " > ";
debugVect(cipEntries[i]);
}
#endif
}
}