C++ RDKit::SubstructMatch Result Different from the Python Implementation

[NiBoyang]

I used the C++ implementation to match a substructure:

#include <iostream>
#include <GraphMol/MolOps.h>
#include <GraphMol/SmilesParse/SmilesParse.h>
#include <GraphMol/Substruct/SubstructMatch.h>

int main()
{
    RDKit::ROMol *mol1 = RDKit::SmilesToMol( "CCOc1cc2c(cc1OCCCN)-c1c(c(Nc3cccc(C(=O)CC)n3)nn1C)C2" );
    RDKit::RWMol *patt = RDKit::SmartsToMol( "[!$(*#*)&!D1]-&!@[!$(*#*)&!D1]" );
    auto testMolWithH = RDKit::MolOps::addHs(*mol1);


    std::vector<RDKit::MatchVectType> hits_vect;

    if ( RDKit::SubstructMatch( *mol1 , *patt , hits_vect ) ) 
    {
        for( size_t i = 0 ; i < hits_vect.size() ; ++i ) 
        {
            std::cout << "Match " << i + 1 << " : ";
            for( size_t j = 0 ; j < hits_vect[i].size() ; ++j ) 
            {
                std::cout << "(" << hits_vect[i][j].first << ","
                    << hits_vect[i][j].second << ")";
            }
        std::cout << std::endl;
        }
    }
}

The C++ result is:

Match 1 : (0,1)(1,2)
Match 2 : (0,2)(1,3)
Match 3 : (0,8)(1,9)
Match 4 : (0,9)(1,10)
Match 5 : (0,10)(1,11)
Match 6 : (0,11)(1,12)
Match 7 : (0,16)(1,17)
Match 8 : (0,17)(1,18)
Match 9 : (0,22)(1,23)
Match 10 : (0,23)(1,25)

The Python Implementation:

from rdkit import Chem

mol = Chem.MolFromSmiles('CCOc1cc2c(cc1OCCCN)-c1c(c(Nc3cccc(C(=O)CC)n3)nn1C)C2')
mol = Chem.AddHs(mol)
rotatable_bond_from_smarts = Chem.MolFromSmarts("[!$(*#*)&!D1]-&!@[!$(*#*)&!D1]")
mol.GetSubstructMatches(rotatable_bond_from_smarts)

The Python result is:

((0, 1),
 (1, 2),
 (2, 3),
 (8, 9),
 (9, 10),
 (10, 11),
 (11, 12),
 (12, 13),
 (16, 17),
 (17, 18),
 (22, 23),
 (23, 25),
 (25, 26),
 (29, 30))

(0,1), (12, 13), (25, 26), (23, 25) were missing from the C++ result.

Any ideas on the reason for this difference? Thanks in advance!

[DavidACosgrove]

Hi @NiBoyang,
I think that's because in your C++ code you're matching against mol1 (without hydrogens) and in the Python you're matching against mol after you've added hydrogens. I suspect that if you use the testMolWithH molecule in your C++ code you will get the same answer in both cases.
Dave