int countAtomElec(const Atom *at) {
PRECONDITION(at, "bad atom");
// default valence :
int dv = PeriodicTable::getTable()->getDefaultValence(at->getAtomicNum());
if (dv <= 1) {
// univalent elements can't be either aromatic or conjugated
return -1;
}
// total atom degree:
int degree = at->getDegree() + at->getTotalNumHs();
ROMol::OEDGE_ITER beg, end;
boost::tie(beg, end) = at->getOwningMol().getAtomBonds(at);
while (beg != end) {
BOND_SPTR bond = at->getOwningMol()[*beg];
if (bond->getBondType() == Bond::UNSPECIFIED // query bonds should not
// contribute; this was github
// issue #443
||
bond->getBondType() == Bond::ZERO)
--degree;
++beg;
}
// if we are more than 3 coordinated we should not be aromatic
if (degree > 3) {
return -1;
}
// number of lone pair electrons = (outer shell elecs) - (default valence)
int nlp;
nlp = PeriodicTable::getTable()->getNouterElecs(at->getAtomicNum()) - dv;
// subtract the charge to get the true number of lone pair electrons:
nlp = std::max(nlp - at->getFormalCharge(), 0);
int nRadicals = at->getNumRadicalElectrons();
// num electrons available for donation into the pi system:
int res = (dv - degree) + nlp - nRadicals;
if (res > 1) {
// if we have an incident bond with order higher than 2,
// (e.g. triple or higher), we only want to return 1 electron
// we detect this using the total unsaturation, because we
// know that there aren't multiple unsaturations (detected
// above in isAtomCandForArom())
int nUnsaturations = at->getExplicitValence() - at->getDegree();
if (nUnsaturations > 1) {
res = 1;
}
}
return res;
}
// 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;
}
}
bool bondCompat(const BOND_SPTR &b1, const BOND_SPTR &b2,
bool useQueryQueryMatches) {
PRECONDITION(b1, "bad bond");
PRECONDITION(b2, "bad bond");
bool res;
if (useQueryQueryMatches && b1->hasQuery() && b2->hasQuery()) {
res = static_cast<QueryBond *>(b1.get())->QueryMatch(
static_cast<QueryBond *>(b2.get()));
} else {
res = b1->Match(b2);
}
if (res && b1->getBondType() == Bond::DATIVE &&
b2->getBondType() == Bond::DATIVE) {
// for dative bonds we need to make sure that the direction also matches:
if (!b1->getBeginAtom()->Match(b1->getBeginAtom()) ||
!b1->getEndAtom()->Match(b2->getEndAtom())) {
res = false;
}
}
// std::cout << "\t\tbondCompat: "<< b1->getIdx() << "-" << b2->getIdx() << ":
// " << res << std::endl;
return res;
}
// finds cycles
void dfsFindCycles(ROMol &mol,int atomIdx,int inBondIdx,
std::vector<AtomColors> &colors,
const UINT_VECT &ranks,
INT_VECT &atomOrders,
VECT_INT_VECT &atomRingClosures,
const boost::dynamic_bitset<> *bondsInPlay,
const std::vector<std::string> *bondSymbols
){
Atom *atom = mol.getAtomWithIdx(atomIdx);
atomOrders.push_back(atomIdx);
colors[atomIdx] = GREY_NODE;
// ---------------------
//
// Build the list of possible destinations from here
//
// ---------------------
std::vector< PossibleType > possibles;
possibles.resize(0);
ROMol::OBOND_ITER_PAIR bondsPair = mol.getAtomBonds(atom);
possibles.reserve(bondsPair.second-bondsPair.first);
while(bondsPair.first != bondsPair.second){
BOND_SPTR theBond = mol[*(bondsPair.first)];
bondsPair.first++;
if(bondsInPlay && !(*bondsInPlay)[theBond->getIdx()]) continue;
if(inBondIdx<0 || theBond->getIdx() != static_cast<unsigned int>(inBondIdx)){
int otherIdx = theBond->getOtherAtomIdx(atomIdx);
long rank=ranks[otherIdx];
// ---------------------
//
// things are a bit more complicated if we are sitting on a
// ring atom. we would like to traverse first to the
// ring-closure atoms, then to atoms outside the ring first,
// then to atoms in the ring that haven't already been visited
// (non-ring-closure atoms).
//
// Here's how the black magic works:
// - non-ring atom neighbors have their original ranks
// - ring atom neighbors have this added to their ranks:
// (MAX_BONDTYPE - bondOrder)*MAX_NATOMS*MAX_NATOMS
// - ring-closure neighbors lose a factor of:
// (MAX_BONDTYPE+1)*MAX_NATOMS*MAX_NATOMS
//
// This tactic biases us to traverse to non-ring neighbors first,
// original ordering if bond orders are all equal... crafty, neh?
//
// ---------------------
if( colors[otherIdx] == GREY_NODE ) {
rank -= static_cast<int>(MAX_BONDTYPE+1) *
MAX_NATOMS*MAX_NATOMS;
if(!bondSymbols){
rank += static_cast<int>(MAX_BONDTYPE - theBond->getBondType()) *
MAX_NATOMS;
} else {
const std::string &symb=(*bondSymbols)[theBond->getIdx()];
boost::uint32_t hsh=gboost::hash_range(symb.begin(),symb.end());
rank += (hsh%MAX_NATOMS) * MAX_NATOMS;
}
} else if( theBond->getOwningMol().getRingInfo()->numBondRings(theBond->getIdx()) ){
if(!bondSymbols){
rank += static_cast<int>(MAX_BONDTYPE - theBond->getBondType()) *
MAX_NATOMS*MAX_NATOMS;
} else {
const std::string &symb=(*bondSymbols)[theBond->getIdx()];
boost::uint32_t hsh=gboost::hash_range(symb.begin(),symb.end());
rank += (hsh%MAX_NATOMS)*MAX_NATOMS*MAX_NATOMS;
}
}
//std::cerr<<"aIdx: "<< atomIdx <<" p: "<<otherIdx<<" Rank: "<<ranks[otherIdx] <<" "<<colors[otherIdx]<<" "<<theBond->getBondType()<<" "<<rank<<std::endl;
possibles.push_back(PossibleType(rank,otherIdx,theBond.get()));
}
}
// ---------------------
//
// Sort on ranks
//
// ---------------------
std::sort(possibles.begin(),possibles.end(),_possibleCompare());
// ---------------------
//
// Now work the children
//
// ---------------------
for(std::vector<PossibleType>::iterator possiblesIt=possibles.begin();
possiblesIt!=possibles.end();
possiblesIt++){
int possibleIdx = possiblesIt->get<1>();
Bond *bond = possiblesIt->get<2>();
Atom *otherAtom=mol.getAtomWithIdx(possibleIdx);
switch(colors[possibleIdx]){
case WHITE_NODE:
// -----
// we haven't seen this node at all before, traverse
// -----
dfsFindCycles(mol,possibleIdx,bond->getIdx(),colors,
ranks,atomOrders,
atomRingClosures,
bondsInPlay,bondSymbols);
break;
case GREY_NODE:
// -----
// we've seen this, but haven't finished it (we're finishing a ring)
// -----
atomRingClosures[possibleIdx].push_back(bond->getIdx());
atomRingClosures[atomIdx].push_back(bond->getIdx());
break;
default:
// -----
// this node has been finished. don't do anything.
// -----
break;
}
}
colors[atomIdx] = BLACK_NODE;
}
RWMOL_SPTR initProduct(const ROMOL_SPTR prodTemplateSptr) {
const ROMol *prodTemplate=prodTemplateSptr.get();
RWMol *res=new RWMol();
// --------- --------- --------- --------- --------- ---------
// Initialize by making a copy of the product template as a normal molecule.
// NOTE that we can't just use a normal copy because we do not want to end up
// with query atoms or bonds in the product.
// copy in the atoms:
ROMol::ATOM_ITER_PAIR atItP = prodTemplate->getVertices();
while(atItP.first != atItP.second ) {
Atom *oAtom=(*prodTemplate)[*(atItP.first++)].get();
Atom *newAtom=new Atom(*oAtom);
res->addAtom(newAtom,false,true);
int mapNum;
if(newAtom->getPropIfPresent(common_properties::molAtomMapNumber, mapNum)) {
// set bookmarks for the mapped atoms:
res->setAtomBookmark(newAtom,mapNum);
// now clear the molAtomMapNumber property so that it doesn't
// end up in the products (this was bug 3140490):
newAtom->clearProp(common_properties::molAtomMapNumber);
}
newAtom->setChiralTag(Atom::CHI_UNSPECIFIED);
// if the product-template atom has the inversion flag set
// to 4 (=SET), then bring its stereochem over, otherwise we'll
// ignore it:
int iFlag;
if(oAtom->getPropIfPresent(common_properties::molInversionFlag, iFlag)) {
if(iFlag==4) newAtom->setChiralTag(oAtom->getChiralTag());
}
// check for properties we need to set:
int val;
if(newAtom->getPropIfPresent(common_properties::_QueryFormalCharge, val)) {
newAtom->setFormalCharge(val);
}
if(newAtom->getPropIfPresent(common_properties::_QueryHCount, val)) {
newAtom->setNumExplicitHs(val);
}
if(newAtom->getPropIfPresent(common_properties::_QueryMass, val)) {
// FIX: technically should do something with this
//newAtom->setMass(val);
}
if(newAtom->getPropIfPresent(common_properties::_QueryIsotope, val)) {
newAtom->setIsotope(val);
}
}
// and the bonds:
ROMol::BOND_ITER_PAIR bondItP = prodTemplate->getEdges();
while(bondItP.first != bondItP.second ) {
const BOND_SPTR oldB=(*prodTemplate)[*(bondItP.first++)];
unsigned int bondIdx;
bondIdx=res->addBond(oldB->getBeginAtomIdx(),oldB->getEndAtomIdx(),oldB->getBondType())-1;
// make sure we don't lose the bond dir information:
Bond *newB=res->getBondWithIdx(bondIdx);
newB->setBondDir(oldB->getBondDir());
// Special case/hack:
// The product has been processed by the SMARTS parser.
// The SMARTS parser tags unspecified bonds as single, but then adds
// a query so that they match single or double
// This caused Issue 1748846
// http://sourceforge.net/tracker/index.php?func=detail&aid=1748846&group_id=160139&atid=814650
// We need to fix that little problem now:
if( oldB->hasQuery()) {
// remember that the product has been processed by the SMARTS parser.
std::string queryDescription=oldB->getQuery()->getDescription();
if(queryDescription=="BondOr" &&
oldB->getBondType()==Bond::SINGLE) {
// We need to fix that little problem now:
if(newB->getBeginAtom()->getIsAromatic() && newB->getEndAtom()->getIsAromatic()) {
newB->setBondType(Bond::AROMATIC);
newB->setIsAromatic(true);
} else {
newB->setBondType(Bond::SINGLE);
newB->setIsAromatic(false);
}
} else if(queryDescription=="BondNull") {
newB->setProp(common_properties::NullBond,1);
}
}
}
return RWMOL_SPTR(res);
} // end of initProduct()
static void addResult(std::vector<std::pair<ROMOL_SPTR, ROMOL_SPTR> >&
res, // const SignatureVector& resSignature,
const ROMol& mol,
const BondVector_t& bonds_selected, size_t maxCuts) {
#ifdef _DEBUG
std::cout << res.size() + 1 << ": ";
#endif
RWMol em(mol);
// loop through the bonds to delete. == deleteBonds()
unsigned isotope = 0;
std::map<unsigned, unsigned> isotope_track;
for (size_t i = 0; i < bonds_selected.size(); i++) {
#ifdef _DEBUG
{
std::string symbol =
em.getAtomWithIdx(bonds_selected[i].first)->getSymbol();
int label = 0;
em.getAtomWithIdx(bonds_selected[i].first)
->getPropIfPresent(common_properties::molAtomMapNumber, label);
char a1[32];
if (0 == label)
sprintf(a1, "\'%s\'", symbol.c_str(), label);
else
sprintf(a1, "\'%s:%u\'", symbol.c_str(), label);
symbol = em.getAtomWithIdx(bonds_selected[i].second)->getSymbol();
label = 0;
em.getAtomWithIdx(bonds_selected[i].second)
->getPropIfPresent(common_properties::molAtomMapNumber, label);
char a2[32];
if (0 == label)
sprintf(a2, "\'%s\'", symbol.c_str(), label);
else
sprintf(a2, "\'%s:%u\'", symbol.c_str(), label);
std::cout << "(" << bonds_selected[i].first << a1 << ","
<< bonds_selected[i].second << a2 << ") ";
}
#endif
isotope += 1;
// remove the bond
em.removeBond(bonds_selected[i].first, bonds_selected[i].second);
// now add attachement points and set attachment point lables
Atom* a = new Atom(0);
a->setProp(common_properties::molAtomMapNumber, (int)isotope);
unsigned newAtomA = em.addAtom(a, true, true);
em.addBond(bonds_selected[i].first, newAtomA, Bond::SINGLE);
a = new Atom(0);
a->setProp(common_properties::molAtomMapNumber, (int)isotope);
unsigned newAtomB = em.addAtom(a, true, true);
em.addBond(bonds_selected[i].second, newAtomB, Bond::SINGLE);
// keep track of where to put isotopes
isotope_track[newAtomA] = isotope;
isotope_track[newAtomB] = isotope;
}
#ifdef _DEBUG
std::cout << "\n";
#endif
RWMOL_SPTR core, side_chains; // core & side_chains output molecules
if (isotope == 1) {
side_chains = RWMOL_SPTR(new RWMol(em)); // output = '%s,%s,,%s.%s'
// DEBUG PRINT
#ifdef _DEBUG
// OK: std::cout<<res.size()+1<<" isotope="<< isotope <<","<<
// MolToSmiles(*side_chains, true) <<"\n";
#endif
} else if (isotope >= 2) {
std::vector<std::vector<int> > frags;
unsigned int nFrags = MolOps::getMolFrags(em, frags);
//#check if its a valid triple or bigger cut. matchObj = re.search(
//'\*.*\*.*\*', f)
// check if exists a fragment with maxCut connection points (*.. *.. *)
if (isotope >= 3) {
bool valid = false;
for (size_t i = 0; i < nFrags; i++) {
unsigned nLabels = 0;
for (size_t ai = 0; ai < frags[i].size(); ai++) {
if (isotope_track.end() !=
isotope_track.find(frags[i][ai])) // new added atom
++nLabels; // found connection point
}
if (nLabels >=
maxCuts) { // looks like it should be selected as core ! ??????
valid = true;
break;
}
}
if (!valid) {
#ifdef _DEBUG
std::cout << "isotope>=3: invalid fragments. fragment with maxCut "
"connection points not found"
<< "\n";
#endif
return;
}
}
size_t iCore = std::numeric_limits<size_t>::max();
side_chains = RWMOL_SPTR(new RWMol);
std::map<unsigned, unsigned>
visitedBonds; // key is bond index in source molecule
unsigned maxAttachments = 0;
for (size_t i = 0; i < frags.size(); i++) {
unsigned nAttachments = 0;
for (size_t ai = 0; ai < frags[i].size(); ai++) {
if (isotope_track.end() !=
isotope_track.find(
frags[i][ai])) // == if(a->hasProp("molAtomMapNumber"))
++nAttachments;
}
if (maxAttachments < nAttachments) maxAttachments = nAttachments;
if (1 == nAttachments) { // build side-chain set of molecules from
// selected fragment
std::map<unsigned, unsigned>
newAtomMap; // key is atom index in source molecule
for (size_t ai = 0; ai < frags[i].size(); ai++) {
Atom* a = em.getAtomWithIdx(frags[i][ai]);
newAtomMap[frags[i][ai]] =
side_chains->addAtom(a->copy(), true, true);
}
// add all bonds from this fragment
for (size_t ai = 0; ai < frags[i].size(); ai++) {
Atom* a = em.getAtomWithIdx(frags[i][ai]);
ROMol::OEDGE_ITER beg, end;
for (boost::tie(beg, end) = em.getAtomBonds(a); beg != end; ++beg) {
const BOND_SPTR bond = em[*beg];
if (newAtomMap.end() == newAtomMap.find(bond->getBeginAtomIdx()) ||
newAtomMap.end() == newAtomMap.find(bond->getEndAtomIdx()) ||
visitedBonds.end() != visitedBonds.find(bond->getIdx()))
continue;
unsigned ai1 = newAtomMap[bond->getBeginAtomIdx()];
unsigned ai2 = newAtomMap[bond->getEndAtomIdx()];
unsigned bi = side_chains->addBond(ai1, ai2, bond->getBondType());
visitedBonds[bond->getIdx()] = bi;
}
}
} else { // select the core fragment
// DEBUG PRINT
#ifdef _DEBUG
if (iCore != -1)
std::cout << "Next CORE found. iCore=" << iCore << " New i=" << i
<< " nAttachments=" << nAttachments << "\n";
#endif
if (nAttachments >= maxAttachments) // Choose a fragment with maximal
// number of connection points as a
// core
iCore = i;
}
}
// build core molecule from selected fragment
if (iCore != std::numeric_limits<size_t>::max()) {
core = RWMOL_SPTR(new RWMol);
visitedBonds.clear();
std::map<unsigned, unsigned>
newAtomMap; // key is atom index in source molecule
for (size_t i = 0; i < frags[iCore].size(); i++) {
unsigned ai = frags[iCore][i];
Atom* a = em.getAtomWithIdx(ai);
newAtomMap[ai] = core->addAtom(a->copy(), true, true);
}
// add all bonds from this fragment
for (size_t ai = 0; ai < frags[iCore].size(); ai++) {
Atom* a = em.getAtomWithIdx(frags[iCore][ai]);
ROMol::OEDGE_ITER beg, end;
for (boost::tie(beg, end) = em.getAtomBonds(a); beg != end; ++beg) {
const BOND_SPTR bond = em[*beg];
if (newAtomMap.end() == newAtomMap.find(bond->getBeginAtomIdx()) ||
newAtomMap.end() == newAtomMap.find(bond->getEndAtomIdx()) ||
visitedBonds.end() != visitedBonds.find(bond->getIdx()))
continue;
unsigned ai1 = newAtomMap[bond->getBeginAtomIdx()];
unsigned ai2 = newAtomMap[bond->getEndAtomIdx()];
unsigned bi = core->addBond(ai1, ai2, bond->getBondType());
visitedBonds[bond->getIdx()] = bi;
}
}
// DEBUG PRINT
#ifdef _DEBUG
// std::cout<<res.size()+1<<" isotope="<< isotope <<" "<< MolToSmiles(*core,
// true)<<", "<<MolToSmiles(*side_chains, true)<<"\n";
#endif
} // iCore != -1
}
// check for duplicates:
bool resFound = false;
size_t ri = 0;
for (ri = 0; ri < res.size(); ri++) {
const std::pair<ROMOL_SPTR, ROMOL_SPTR>& r = res[ri];
if (side_chains->getNumAtoms() == r.second->getNumAtoms() &&
side_chains->getNumBonds() == r.second->getNumBonds() &&
((NULL == core.get() && NULL == r.first.get()) ||
(NULL != core.get() && NULL != r.first.get() &&
core->getNumAtoms() == r.first->getNumAtoms() &&
core->getNumBonds() == r.first->getNumBonds()))) {
// ToDo accurate check:
// 1. compare hash code
if (computeMorganCodeHash(*side_chains) ==
computeMorganCodeHash(*r.second) &&
(NULL == core ||
computeMorganCodeHash(*core) == computeMorganCodeHash(*r.first))) {
// 2. final check to exclude hash collisions
// We decided that it does not neccessary to implement
resFound = true;
break;
}
}
}
if (!resFound) {
//std::cerr << "**********************" << std::endl;
// From rfrag.py
// now change the labels on sidechains and core
// to get the new labels, cansmi the dot-disconnected side chains
// the first fragment in the side chains has attachment label 1, 2nd: 2, 3rd: 3
// then change the labels accordingly in the core
std::map<unsigned int, int> canonicalAtomMaps;
if( side_chains.get() ) {
RWMol tmp_side_chain(*(side_chains.get()));
std::vector<int> oldMaps(tmp_side_chain.getNumAtoms(), 0);
// clear atom labels (they are used in canonicalization)
// and move them to dummy storage
for (ROMol::AtomIterator at = tmp_side_chain.beginAtoms(); at != tmp_side_chain.endAtoms();
++at) {
int label = 0;
if ((*at)->getPropIfPresent(common_properties::molAtomMapNumber, label) ) {
(*at)->clearProp(common_properties::molAtomMapNumber);
oldMaps[(*at)->getIdx()] = label;
}
}
const bool doIsomericSmiles = true; // should this be false???
std::string smiles = MolToSmiles(tmp_side_chain, doIsomericSmiles);
//std::cerr << "smiles: " << smiles << std::endl;
// Get the canonical output order and use it to remap
// the atom maps int the side chains
// these will get reapplied to the core (if there is a core)
const std::vector<unsigned int> &ranks = tmp_side_chain.getProp<
std::vector<unsigned int> >(
common_properties::_smilesAtomOutputOrder);
std::vector<std::pair<unsigned int, int> > rankedAtoms;
for(size_t idx=0;idx<ranks.size();++idx) {
unsigned int atom_idx = ranks[idx];
if(oldMaps[atom_idx] >0) {
const int label = oldMaps[atom_idx];
//std::cerr << "atom_idx: " << atom_idx << " rank: " << ranks[atom_idx] <<
// " molAtomMapNumber: " << label << std::endl;
rankedAtoms.push_back(std::make_pair(idx, label));
}
}
std::sort(rankedAtoms.begin(), rankedAtoms.end());
int nextMap = 0;
for(size_t i=0;i<rankedAtoms.size();++i) {
if(canonicalAtomMaps.find(rankedAtoms[i].second) == canonicalAtomMaps.end()) {
//std::cerr << "Remapping: " << rankedAtoms[i].second << " " << " to " << (i+1) <<
// std::endl;
canonicalAtomMaps[rankedAtoms[i].second] = ++nextMap;
}
}
}
//std::cerr << "======== Remap core " << std::endl;
if( core.get() ) { // remap core if it exists
for (ROMol::AtomIterator at = core->beginAtoms(); at != core->endAtoms();
++at) {
int label = 0;
if ((*at)->getPropIfPresent(common_properties::molAtomMapNumber, label) ) {
//std::cerr << "remapping core: " << label << " :" << canonicalAtomMaps[label] <<
// std::endl;
(*at)->setProp(common_properties::molAtomMapNumber, canonicalAtomMaps[label]);
}
}
}
//std::cerr << "======== Remap side-chain " << std::endl;
for (ROMol::AtomIterator at = side_chains->beginAtoms(); at != side_chains->endAtoms();
++at) {
int label = 0;
if ((*at)->getPropIfPresent(common_properties::molAtomMapNumber, label) ) {
//std::cerr << "remapping side chain: " << label << " :" <<
// canonicalAtomMaps[label] << std::endl;
(*at)->setProp(common_properties::molAtomMapNumber, canonicalAtomMaps[label]);
}
}
res.push_back(std::pair<ROMOL_SPTR, ROMOL_SPTR>(core, side_chains)); //
}
#ifdef _DEBUG
else
std::cout << res.size() + 1 << " --- DUPLICATE Result FOUND --- ri=" << ri
<< "\n";
#endif
}
static void addResult(std::vector<std::pair<ROMOL_SPTR, ROMOL_SPTR> >&
res, // const SignatureVector& resSignature,
const ROMol& mol,
const BondVector_t& bonds_selected, size_t maxCuts) {
#ifdef _DEBUG
std::cout << res.size() + 1 << ": ";
#endif
RWMol em(mol);
// loop through the bonds to delete. == deleteBonds()
unsigned isotope = 0;
std::map<unsigned, unsigned> isotope_track;
for (size_t i = 0; i < bonds_selected.size(); i++) {
#ifdef _DEBUG
{
std::string symbol =
em.getAtomWithIdx(bonds_selected[i].first)->getSymbol();
int label = 0;
em.getAtomWithIdx(bonds_selected[i].first)
->getPropIfPresent(common_properties::molAtomMapNumber, label);
char a1[32];
if (0 == label)
sprintf(a1, "\'%s\'", symbol.c_str(), label);
else
sprintf(a1, "\'%s:%u\'", symbol.c_str(), label);
symbol = em.getAtomWithIdx(bonds_selected[i].second)->getSymbol();
label = 0;
em.getAtomWithIdx(bonds_selected[i].second)
->getPropIfPresent(common_properties::molAtomMapNumber, label);
char a2[32];
if (0 == label)
sprintf(a2, "\'%s\'", symbol.c_str(), label);
else
sprintf(a2, "\'%s:%u\'", symbol.c_str(), label);
std::cout << "(" << bonds_selected[i].first << a1 << ","
<< bonds_selected[i].second << a2 << ") ";
}
#endif
isotope += 1;
// remove the bond
em.removeBond(bonds_selected[i].first, bonds_selected[i].second);
// now add attachement points and set attachment point lables
Atom* a = new Atom(0);
a->setProp(common_properties::molAtomMapNumber, (int)isotope);
unsigned newAtomA = em.addAtom(a, true, true);
em.addBond(bonds_selected[i].first, newAtomA, Bond::SINGLE);
a = new Atom(0);
a->setProp(common_properties::molAtomMapNumber, (int)isotope);
unsigned newAtomB = em.addAtom(a, true, true);
em.addBond(bonds_selected[i].second, newAtomB, Bond::SINGLE);
// keep track of where to put isotopes
isotope_track[newAtomA] = isotope;
isotope_track[newAtomB] = isotope;
}
#ifdef _DEBUG
std::cout << "\n";
#endif
RWMOL_SPTR core, side_chains; // core & side_chains output molecules
if (isotope == 1) {
side_chains = RWMOL_SPTR(new RWMol(em)); // output = '%s,%s,,%s.%s'
// DEBUG PRINT
#ifdef _DEBUG
// OK: std::cout<<res.size()+1<<" isotope="<< isotope <<","<<
// MolToSmiles(*side_chains, true) <<"\n";
#endif
} else if (isotope >= 2) {
std::vector<std::vector<int> > frags;
unsigned int nFrags = MolOps::getMolFrags(em, frags);
//#check if its a valid triple or bigger cut. matchObj = re.search(
//'\*.*\*.*\*', f)
// check if exists a fragment with maxCut connection points (*.. *.. *)
if (isotope >= 3) {
bool valid = false;
for (size_t i = 0; i < nFrags; i++) {
unsigned nLabels = 0;
for (size_t ai = 0; ai < frags[i].size(); ai++) {
if (isotope_track.end() !=
isotope_track.find(frags[i][ai])) // new added atom
++nLabels; // found connection point
}
if (nLabels >=
maxCuts) { // looks like it should be selected as core ! ??????
valid = true;
break;
}
}
if (!valid) {
#ifdef _DEBUG
std::cout << "isotope>=3: invalid fragments. fragment with maxCut "
"connection points not found"
<< "\n";
#endif
return;
}
}
size_t iCore = std::numeric_limits<size_t>::max();
side_chains = RWMOL_SPTR(new RWMol);
std::map<unsigned, unsigned>
visitedBonds; // key is bond index in source molecule
unsigned maxAttachments = 0;
for (size_t i = 0; i < frags.size(); i++) {
unsigned nAttachments = 0;
for (size_t ai = 0; ai < frags[i].size(); ai++) {
if (isotope_track.end() !=
isotope_track.find(
frags[i][ai])) // == if(a->hasProp("molAtomMapNumber"))
++nAttachments;
}
if (maxAttachments < nAttachments) maxAttachments = nAttachments;
if (1 == nAttachments) { // build side-chain set of molecules from
// selected fragment
std::map<unsigned, unsigned>
newAtomMap; // key is atom index in source molecule
for (size_t ai = 0; ai < frags[i].size(); ai++) {
Atom* a = em.getAtomWithIdx(frags[i][ai]);
newAtomMap[frags[i][ai]] =
side_chains->addAtom(a->copy(), true, true);
}
// add all bonds from this fragment
for (size_t ai = 0; ai < frags[i].size(); ai++) {
Atom* a = em.getAtomWithIdx(frags[i][ai]);
ROMol::OEDGE_ITER beg, end;
for (boost::tie(beg, end) = em.getAtomBonds(a); beg != end; ++beg) {
const BOND_SPTR bond = em[*beg];
if (newAtomMap.end() == newAtomMap.find(bond->getBeginAtomIdx()) ||
newAtomMap.end() == newAtomMap.find(bond->getEndAtomIdx()) ||
visitedBonds.end() != visitedBonds.find(bond->getIdx()))
continue;
unsigned ai1 = newAtomMap[bond->getBeginAtomIdx()];
unsigned ai2 = newAtomMap[bond->getEndAtomIdx()];
unsigned bi = side_chains->addBond(ai1, ai2, bond->getBondType());
visitedBonds[bond->getIdx()] = bi;
}
}
} else { // select the core fragment
// DEBUG PRINT
#ifdef _DEBUG
if (iCore != -1)
std::cout << "Next CORE found. iCore=" << iCore << " New i=" << i
<< " nAttachments=" << nAttachments << "\n";
#endif
if (nAttachments >= maxAttachments) // Choose a fragment with maximal
// number of connection points as a
// core
iCore = i;
}
}
// build core molecule from selected fragment
if (iCore != std::numeric_limits<size_t>::max()) {
core = RWMOL_SPTR(new RWMol);
visitedBonds.clear();
std::map<unsigned, unsigned>
newAtomMap; // key is atom index in source molecule
for (size_t i = 0; i < frags[iCore].size(); i++) {
unsigned ai = frags[iCore][i];
Atom* a = em.getAtomWithIdx(ai);
newAtomMap[ai] = core->addAtom(a->copy(), true, true);
}
// add all bonds from this fragment
for (size_t ai = 0; ai < frags[iCore].size(); ai++) {
Atom* a = em.getAtomWithIdx(frags[iCore][ai]);
ROMol::OEDGE_ITER beg, end;
for (boost::tie(beg, end) = em.getAtomBonds(a); beg != end; ++beg) {
const BOND_SPTR bond = em[*beg];
if (newAtomMap.end() == newAtomMap.find(bond->getBeginAtomIdx()) ||
newAtomMap.end() == newAtomMap.find(bond->getEndAtomIdx()) ||
visitedBonds.end() != visitedBonds.find(bond->getIdx()))
continue;
unsigned ai1 = newAtomMap[bond->getBeginAtomIdx()];
unsigned ai2 = newAtomMap[bond->getEndAtomIdx()];
unsigned bi = core->addBond(ai1, ai2, bond->getBondType());
visitedBonds[bond->getIdx()] = bi;
}
}
// DEBUG PRINT
#ifdef _DEBUG
// std::cout<<res.size()+1<<" isotope="<< isotope <<" "<< MolToSmiles(*core,
// true)<<", "<<MolToSmiles(*side_chains, true)<<"\n";
#endif
} // iCore != -1
}
// check for dublicates:
bool resFound = false;
size_t ri = 0;
for (ri = 0; ri < res.size(); ri++) {
const std::pair<ROMOL_SPTR, ROMOL_SPTR>& r = res[ri];
if (side_chains->getNumAtoms() == r.second->getNumAtoms() &&
side_chains->getNumBonds() == r.second->getNumBonds() &&
((NULL == core.get() && NULL == r.first.get()) ||
(NULL != core.get() && NULL != r.first.get() &&
core->getNumAtoms() == r.first->getNumAtoms() &&
core->getNumBonds() == r.first->getNumBonds()))) {
// ToDo accurate check:
// 1. compare hash code
if (computeMorganCodeHash(*side_chains) ==
computeMorganCodeHash(*r.second) &&
(NULL == core ||
computeMorganCodeHash(*core) == computeMorganCodeHash(*r.first))) {
// 2. final check to exclude hash collisions
// We decided that it does not neccessary to implement
resFound = true;
break;
}
}
}
if (!resFound)
res.push_back(std::pair<ROMOL_SPTR, ROMOL_SPTR>(core, side_chains)); //
#ifdef _DEBUG
else
std::cout << res.size() + 1 << " --- DUPLICATE Result FOUND --- ri=" << ri
<< "\n";
#endif
}
void canonicalDFSTraversal(ROMol &mol,int atomIdx,int inBondIdx,
std::vector<AtomColors> &colors,
VECT_INT_VECT &cycles,
INT_VECT &ranks,
INT_VECT &cyclesAvailable,
MolStack &molStack,
INT_VECT &atomOrders,
INT_VECT &bondVisitOrders,
VECT_INT_VECT &atomRingClosures,
std::vector<INT_LIST> &atomTraversalBondOrder,
const boost::dynamic_bitset<> *bondsInPlay,
const std::vector<std::string> *bondSymbols
){
PRECONDITION(colors.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(ranks.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(atomOrders.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(bondVisitOrders.size()>=mol.getNumBonds(),"vector too small");
PRECONDITION(atomRingClosures.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(atomTraversalBondOrder.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(!bondsInPlay || bondsInPlay->size()>=mol.getNumBonds(),"bondsInPlay too small");
PRECONDITION(!bondSymbols || bondSymbols->size()>=mol.getNumBonds(),"bondSymbols too small");
int nAttached=0;
Atom *atom = mol.getAtomWithIdx(atomIdx);
INT_LIST directTravList,cycleEndList;
molStack.push_back(MolStackElem(atom));
atomOrders[atom->getIdx()] = molStack.size();
colors[atomIdx] = GREY_NODE;
// ---------------------
//
// Build the list of possible destinations from here
//
// ---------------------
std::vector< PossibleType > possibles;
possibles.resize(0);
ROMol::OBOND_ITER_PAIR bondsPair = mol.getAtomBonds(atom);
possibles.reserve(bondsPair.second-bondsPair.first);
while(bondsPair.first != bondsPair.second){
BOND_SPTR theBond = mol[*(bondsPair.first)];
bondsPair.first++;
if(bondsInPlay && !(*bondsInPlay)[theBond->getIdx()]) continue;
if(inBondIdx<0 || theBond->getIdx() != static_cast<unsigned int>(inBondIdx)){
int otherIdx = theBond->getOtherAtomIdx(atomIdx);
long rank=ranks[otherIdx];
// ---------------------
//
// things are a bit more complicated if we are sitting on a
// ring atom we would like to traverse first to the
// ring-closure atoms, then to atoms outside the ring first,
// then to atoms in the ring that haven't already been visited
// (non-ring-closure atoms).
//
// Here's how the black magic works:
// - non-ring atom neighbors have their original ranks
// - ring atom neighbors have this added to their ranks:
// (Bond::OTHER - bondOrder)*MAX_NATOMS*MAX_NATOMS
// - ring-closure neighbors lose a factor of:
// (Bond::OTHER+1)*MAX_NATOMS*MAX_NATOMS
//
// This tactic biases us to traverse to non-ring neighbors first,
// original ordering if bond orders are all equal... crafty, neh?
//
// ---------------------
if( colors[otherIdx] == GREY_NODE ) {
rank -= static_cast<int>(Bond::OTHER+1) *
MAX_NATOMS*MAX_NATOMS;
if(!bondSymbols){
rank += static_cast<int>(Bond::OTHER - theBond->getBondType()) *
MAX_NATOMS;
} else {
const std::string &symb=(*bondSymbols)[theBond->getIdx()];
boost::uint32_t hsh=gboost::hash_range(symb.begin(),symb.end());
rank += (hsh%MAX_NATOMS) * MAX_NATOMS;
}
} else if( theBond->getOwningMol().getRingInfo()->numBondRings(theBond->getIdx()) ){
if(!bondSymbols){
rank += static_cast<int>(Bond::OTHER - theBond->getBondType()) *
MAX_NATOMS*MAX_NATOMS;
} else {
const std::string &symb=(*bondSymbols)[theBond->getIdx()];
boost::uint32_t hsh=gboost::hash_range(symb.begin(),symb.end());
rank += (hsh%MAX_NATOMS)*MAX_NATOMS*MAX_NATOMS;
}
}
possibles.push_back(PossibleType(rank,otherIdx,theBond.get()));
}
}
// ---------------------
//
// Sort on ranks
//
// ---------------------
std::sort(possibles.begin(),possibles.end(),_possibleComp);
// ---------------------
//
// Now work the children
//
// ---------------------
std::vector<MolStack> subStacks;
for(std::vector<PossibleType>::iterator possiblesIt=possibles.begin();
possiblesIt!=possibles.end();
possiblesIt++){
MolStack subStack;
#if 0
int possibleIdx = possiblesIt->second.first;
Bond *bond = possiblesIt->second.second;
#endif
int possibleIdx = possiblesIt->get<1>();
Bond *bond = possiblesIt->get<2>();
Atom *otherAtom=mol.getAtomWithIdx(possibleIdx);
unsigned int lowestRingIdx;
INT_VECT::const_iterator cAIt;
switch(colors[possibleIdx]){
case WHITE_NODE:
// -----
// we haven't seen this node at all before
// -----
// it might have some residual data from earlier calls, clean that up:
if(otherAtom->hasProp("_TraversalBondIndexOrder")){
otherAtom->clearProp("_TraversalBondIndexOrder");
}
directTravList.push_back(bond->getIdx());
subStack.push_back(MolStackElem(bond,atomIdx));
canonicalDFSTraversal(mol,possibleIdx,bond->getIdx(),colors,
cycles,ranks,cyclesAvailable,subStack,
atomOrders,bondVisitOrders,atomRingClosures,atomTraversalBondOrder,
bondsInPlay,bondSymbols);
subStacks.push_back(subStack);
nAttached += 1;
break;
case GREY_NODE:
// -----
// we've seen this, but haven't finished it (we're finishing a ring)
// -----
cycleEndList.push_back(bond->getIdx());
cAIt=std::find(cyclesAvailable.begin(),
cyclesAvailable.end(),1);
if(cAIt==cyclesAvailable.end()){
throw ValueErrorException("Too many rings open at once. SMILES cannot be generated.");
}
lowestRingIdx = cAIt-cyclesAvailable.begin();
cyclesAvailable[lowestRingIdx] = 0;
cycles[possibleIdx].push_back(lowestRingIdx);
++lowestRingIdx;
bond->setProp("_TraversalRingClosureBond",lowestRingIdx);
molStack.push_back(MolStackElem(bond,
atom->getIdx()));
molStack.push_back(MolStackElem(lowestRingIdx));
// we need to add this bond (which closes the ring) to the traversal list for the
// other atom as well:
atomTraversalBondOrder[otherAtom->getIdx()].push_back(bond->getIdx());
atomRingClosures[otherAtom->getIdx()].push_back(bond->getIdx());
break;
default:
// -----
// this node has been finished. don't do anything.
// -----
break;
}
}
INT_VECT &ringClosures=atomRingClosures[atom->getIdx()];
CHECK_INVARIANT(ringClosures.size()==cycles[atomIdx].size(),
"ring closure mismatch");
for(unsigned int i=0;i<ringClosures.size();i++){
int ringIdx=cycles[atomIdx][i];
ringIdx += 1;
molStack.push_back(MolStackElem(ringIdx));
}
cycles[atomIdx].resize(0);
MolStack::const_iterator ciMS;
for(int i=0;i<nAttached;i++){
if(i<nAttached-1){
int branchIdx=0;
if(subStacks[i].begin()->type==MOL_STACK_ATOM){
branchIdx=subStacks[i].begin()->obj.atom->getIdx();
} else if(subStacks[i].begin()->type==MOL_STACK_BOND){
branchIdx=-1*subStacks[i].begin()->obj.bond->getIdx();
} else {
ASSERT_INVARIANT(0,"branch started with something other than an atom or bond");
}
molStack.push_back(MolStackElem("(",branchIdx));
for(ciMS=subStacks[i].begin();ciMS!=subStacks[i].end();ciMS++){
molStack.push_back(*ciMS);
switch(ciMS->type){
case MOL_STACK_ATOM:
atomOrders[ciMS->obj.atom->getIdx()] = molStack.size();
break;
case MOL_STACK_BOND:
bondVisitOrders[ciMS->obj.bond->getIdx()] = molStack.size();
break;
default:
break;
}
}
molStack.push_back(MolStackElem(")",branchIdx));
} else {
for(ciMS=subStacks[i].begin();ciMS!=subStacks[i].end();ciMS++){
molStack.push_back(*ciMS);
switch(ciMS->type){
case MOL_STACK_ATOM:
atomOrders[ciMS->obj.atom->getIdx()] = molStack.size();
break;
case MOL_STACK_BOND:
bondVisitOrders[ciMS->obj.bond->getIdx()] = molStack.size();
break;
default:
break;
}
}
}
}
//std::cerr<<"*****>>>>>> Traversal results for atom: "<<atom->getIdx()<<"> ";
INT_LIST travList;
// first push on the incoming bond:
if(inBondIdx >= 0){
//std::cerr<<" "<<inBondIdx;
travList.push_back(inBondIdx);
}
// ... ring closures that end here:
for(INT_LIST_CI ilci=cycleEndList.begin();ilci!=cycleEndList.end();++ilci){
//std::cerr<<" ["<<*ilci<<"]";
travList.push_back(*ilci);
}
// ... ring closures that start here:
// if(atom->hasProp("_TraversalBondIndexOrder")){
// INT_LIST indirectTravList;
// atom->getProp("_TraversalBondIndexOrder",indirectTravList);
// for(INT_LIST_CI ilci=indirectTravList.begin();ilci!=indirectTravList.end();++ilci){
// //std::cerr<<" ("<<*ilci<<")";
// travList.push_back(*ilci);
// }
// }
BOOST_FOREACH(int ili,atomTraversalBondOrder[atom->getIdx()]){
travList.push_back(ili);
}