1D pharmacophores in P450 models

During the development of SMARTCyp 2 we realized something profound, site-of-metabolism in CYP2D6 and CYP2C9 can be modelled by reactivity and a simple 1-dimensional pharmacophore. Simply by measuring the distance from the atom of interest to the pharmacophore we get a good variable to use for correcting the energy with.

So how come this simple correction works? It is rather well known that cytochromes P450s are quite flexible. A homology model of CYP2D6 has been shown to be more reliable for site-of-metabolism prediction through docking than the crystal structure, indicating a flexible site. The CYP2C9 structure has also been shown to be able to take on multiple conformations. Hence, a structure based model will require many different conformations to give accurate description of the binding, whereas a non-structure based model (i.e. SMARTCyp) simply implicitly includes this flexibility by assuming that there are very few restraints on the rotation of the substrate. SMARTCyp only uses the relative distance (no. of bonds) to the center of the molecule as a small penalty to the reactivity to prioritize sites that are closer to the end of the molecule than the center.


For the implementation of 1D pharmacophores we devised SMARTS strings that match the pharmacophores of CYP2D6 and CYP2C9 that matches the pharmacophores of the isoforms (strings below from SMARTCyp 2.2).

CYP2D6: positively charged amines
$([N][CX3](=[N])[N]) //guanidine like fragment
$([N^3X3H0]([#6^3])([#6^3])[#6^3]),$([N^3X3H1]([#6^3])[#6^3]),$([N^3X3H2][#6^3]) // primary, secondary, tertiary amines bound to only carbon and hydrogen atoms, not next to sp2 carbon
                                

CYP2C9: carboxylic acids and their bioisosteres
$([O]=[C^2][OH1]) // carboxylic acid oxygen
$([O]=[C^2][C^2]=[C^2][OH1]),$([O]=[C^2][c][c][OH1]) // vinylogous carboxylic acids (e.g. ascorbic acid)
$([n]1:[n]:[n]:[n]:[c]1) // tetrazole 1
$([n]1:[n]:[n]:[c]:[n]1) // tetrazole 2
$([O]=[C^2][N][OH1]) // hydroxamic acid
$([O]=[C^2]([N])[N]) // urea
$([O]=[S][OH1]) // sulfinic and sulfonic acids
$([O]=[PD4][OH1]) // phosphate esters and phosphoric acids
$([O]=[S](=[O])(c)[C][C]=[O]),$([O]=[C][C][S](=[O])(=[O])[c]) // sulfones next to phenyls with carbonyl two bonds away
$([O]=[S](=[O])[NH1][C]=[O]),$([O]=[C][NH1][S](=[O])=[O]) // sulfones bound to nitrogen with carbonyl next to it
$([O]=[C^2][NH1][O]),$([O]=[C^2][NH1][C]#[N]) // peptide with oxygen or cyano group next to nitrogen
$([OH1][c]1[n][o,s][c,n][c]1),$([OH1][n]1[n][c,n][c][c]1),$([OH1][n]1[c][n][c][c]1) // alcohol on aromatic five membered ring
$([O]=[C]1[N][C](=O)[O,S][C,N]1) // carbonyl oxygen on almost conjugated five membered ring
$([O]=[C]1[NH1,O][N]=[N,C][N]1) // carbonyl oxygen on fully conjugated five membered ring
$([nD2]1[nD2][c]([S]=[O])[nD2][c]1),$([nD2]1[c]([S]=[O])[nD2][c][nD2]1),$([nD2]1[c]([S]=[O])[nD2][nD2][c]1) // nitrogens in histidine-like 5-ring with sulfoxide/sulfone next to it
$([O]=[SX4](=[O])[NX3]) // sulfonamides

References:
2D6 model
Patrik Rydberg and Lars Olsen "Ligand-Based Site of Metabolism Prediction for Cytochrome P450 2D6" ACS Med. Chem. Lett., 2012, 3, 69-73
2C9 model
Patrik Rydberg and Lars Olsen "Predicting Drug Metabolism by Cytochrome P450 2C9: Comparison with the 2D6 and 3A4 Isoforms" ChemMedChem, 2012, 7, 1202-1209