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
}
}
void checkAndCorrectChiralityOfMatchingAtomsInProduct(
const ROMol &reactant, unsigned reactantAtomIdx, const Atom &reactantAtom,
RWMOL_SPTR product, ReactantProductAtomMapping *mapping) {
for (unsigned i = 0; i < mapping->reactProdAtomMap[reactantAtomIdx].size();
i++) {
unsigned productAtomIdx = mapping->reactProdAtomMap[reactantAtomIdx][i];
Atom *productAtom = product->getAtomWithIdx(productAtomIdx);
if (productAtom->getChiralTag() != Atom::CHI_UNSPECIFIED ||
reactantAtom.getChiralTag() == Atom::CHI_UNSPECIFIED ||
reactantAtom.getChiralTag() == Atom::CHI_OTHER ||
productAtom->hasProp(common_properties::molInversionFlag)) {
continue;
}
// we can only do something sensible here if we have the same number of
// bonds in the reactants and the products:
if (reactantAtom.getDegree() != productAtom->getDegree()) {
continue;
}
unsigned int nUnknown = 0;
INT_LIST pOrder;
ROMol::ADJ_ITER nbrIdx, endNbrs;
boost::tie(nbrIdx, endNbrs) = product->getAtomNeighbors(productAtom);
while (nbrIdx != endNbrs) {
if (mapping->prodReactAtomMap.find(*nbrIdx) ==
mapping->prodReactAtomMap.end() ||
!reactant.getBondBetweenAtoms(reactantAtom.getIdx(),
mapping->prodReactAtomMap[*nbrIdx])) {
++nUnknown;
// if there's more than one bond in the product that doesn't correspond
// to anything in the reactant, we're also doomed
if (nUnknown > 1) break;
// otherwise, add a -1 to the bond order that we'll fill in later
pOrder.push_back(-1);
} else {
const Bond *rBond = reactant.getBondBetweenAtoms(
reactantAtom.getIdx(), mapping->prodReactAtomMap[*nbrIdx]);
CHECK_INVARIANT(rBond, "expected reactant bond not found");
pOrder.push_back(rBond->getIdx());
}
++nbrIdx;
}
if (nUnknown == 1) {
// find the reactant bond that hasn't yet been accounted for:
int unmatchedBond = -1;
boost::tie(nbrIdx, endNbrs) = reactant.getAtomNeighbors(&reactantAtom);
while (nbrIdx != endNbrs) {
const Bond *rBond =
reactant.getBondBetweenAtoms(reactantAtom.getIdx(), *nbrIdx);
if (std::find(pOrder.begin(), pOrder.end(), rBond->getIdx()) ==
pOrder.end()) {
unmatchedBond = rBond->getIdx();
break;
}
++nbrIdx;
}
// what must be true at this point:
// 1) there's a -1 in pOrder that we'll substitute for
// 2) unmatchedBond contains the index of the substitution
auto bPos = std::find(pOrder.begin(), pOrder.end(), -1);
if (unmatchedBond >= 0 && bPos != pOrder.end()) {
*bPos = unmatchedBond;
}
if (std::find(pOrder.begin(), pOrder.end(), -1) == pOrder.end()) {
nUnknown = 0;
}
}
if (!nUnknown) {
productAtom->setChiralTag(reactantAtom.getChiralTag());
int nSwaps = reactantAtom.getPerturbationOrder(pOrder);
if (nSwaps % 2) {
productAtom->invertChirality();
}
}
}
}
//
// Determine bond wedge state
///
Bond::BondDir DetermineBondWedgeState(const Bond *bond,
const INT_MAP_INT &wedgeBonds,
const Conformer *conf) {
PRECONDITION(bond, "no bond");
PRECONDITION(bond->getBondType() == Bond::SINGLE,
"bad bond order for wedging");
const ROMol *mol = &(bond->getOwningMol());
PRECONDITION(mol, "no mol");
Bond::BondDir res = bond->getBondDir();
if (!conf) {
return res;
}
int bid = bond->getIdx();
INT_MAP_INT_CI wbi = wedgeBonds.find(bid);
if (wbi == wedgeBonds.end()) {
return res;
}
unsigned int waid = wbi->second;
Atom *atom, *bondAtom; // = bond->getBeginAtom();
if (bond->getBeginAtom()->getIdx() == waid) {
atom = bond->getBeginAtom();
bondAtom = bond->getEndAtom();
} else {
atom = bond->getEndAtom();
bondAtom = bond->getBeginAtom();
}
Atom::ChiralType chiralType = atom->getChiralTag();
CHECK_INVARIANT(chiralType == Atom::CHI_TETRAHEDRAL_CW ||
chiralType == Atom::CHI_TETRAHEDRAL_CCW,
"");
// if we got this far, we really need to think about it:
INT_LIST neighborBondIndices;
DOUBLE_LIST neighborBondAngles;
RDGeom::Point3D centerLoc, tmpPt;
centerLoc = conf->getAtomPos(atom->getIdx());
tmpPt = conf->getAtomPos(bondAtom->getIdx());
centerLoc.z = 0.0;
tmpPt.z = 0.0;
RDGeom::Point3D refVect = centerLoc.directionVector(tmpPt);
neighborBondIndices.push_back(bond->getIdx());
neighborBondAngles.push_back(0.0);
ROMol::OEDGE_ITER beg, end;
boost::tie(beg, end) = mol->getAtomBonds(atom);
while (beg != end) {
Bond *nbrBond = (*mol)[*beg].get();
Atom *otherAtom = nbrBond->getOtherAtom(atom);
if (nbrBond != bond) {
tmpPt = conf->getAtomPos(otherAtom->getIdx());
tmpPt.z = 0.0;
RDGeom::Point3D tmpVect = centerLoc.directionVector(tmpPt);
double angle = refVect.signedAngleTo(tmpVect);
if (angle < 0.0) angle += 2. * M_PI;
INT_LIST::iterator nbrIt = neighborBondIndices.begin();
DOUBLE_LIST::iterator angleIt = neighborBondAngles.begin();
// find the location of this neighbor in our angle-sorted list
// of neighbors:
while (angleIt != neighborBondAngles.end() && angle > (*angleIt)) {
++angleIt;
++nbrIt;
}
neighborBondAngles.insert(angleIt, angle);
neighborBondIndices.insert(nbrIt, nbrBond->getIdx());
}
++beg;
}
// at this point, neighborBondIndices contains a list of bond
// indices from the central atom. They are arranged starting
// at the reference bond in CCW order (based on the current
// depiction).
int nSwaps = atom->getPerturbationOrder(neighborBondIndices);
// in the case of three-coordinated atoms we may have to worry about
// the location of the implicit hydrogen - Issue 209
// Check if we have one of these situation
//
// 0 1 0 2
// * \*/
// 1 - C - 2 C
//
// here the hydrogen will be between 1 and 2 and we need to add an additional
// swap
if (neighborBondAngles.size() == 3) {
// three coordinated
DOUBLE_LIST::iterator angleIt = neighborBondAngles.begin();
++angleIt; // the first is the 0 (or reference bond - we will ignoire that
double angle1 = (*angleIt);
++angleIt;
double angle2 = (*angleIt);
if (angle2 - angle1 >= (M_PI - 1e-4)) {
// we have the above situation
nSwaps++;
}
}
#ifdef VERBOSE_STEREOCHEM
BOOST_LOG(rdDebugLog) << "--------- " << nSwaps << std::endl;
std::copy(neighborBondIndices.begin(), neighborBondIndices.end(),
std::ostream_iterator<int>(BOOST_LOG(rdDebugLog), " "));
BOOST_LOG(rdDebugLog) << std::endl;
std::copy(neighborBondAngles.begin(), neighborBondAngles.end(),
std::ostream_iterator<double>(BOOST_LOG(rdDebugLog), " "));
BOOST_LOG(rdDebugLog) << std::endl;
#endif
if (chiralType == Atom::CHI_TETRAHEDRAL_CCW) {
if (nSwaps % 2 == 1) { // ^ reverse) {
res = Bond::BEGINDASH;
} else {
res = Bond::BEGINWEDGE;
}
} else {
if (nSwaps % 2 == 1) { // ^ reverse) {
res = Bond::BEGINWEDGE;
} else {
res = Bond::BEGINDASH;
}
}
return res;
}