Wrong MMFF atom types with aromatic rings

[apiserchia]

Given a SMILE containing aromatic rings, MMFF methods use their own aromaticity model therefore changing molecule's aromaticity. In this way I get wrong MMFF94 atom types and energy contributions. I read on the following article "Bringing the MMFF force field to the RDKit: implementation and validation" that

"Since the aromaticity model used by MMFF differs from the one normally used throughout the RDKit, aromaticity has to be re-perceived according to MMFF criteria starting from a kekulized representation of the molecule. Subsequently, atom types are assigned to heavy atoms followed by hydrogens."

Also RDKit Documentation clearly states that

"Please note that the MMFF atom typing code uses its own aromaticity model, so the aromaticity flags of the molecule will be modified after calling MMFF-related methods."

Is there a method/way to retrieve correct MMFF94 atom types and energy contributions from a given SMILE? If yes, what should I do? Am I missing something in the following snippet of Python code?

To Reproduce

from rdkit import Chem
from rdkit.Chem import AllChem

#Thymine
molecule = Chem.MolFromSmiles('O=C1NC(=O)NC=C1C')

moleculewH = Chem.AddHs(molecule)

AllChem.EmbedMolecule(moleculewH)

properties = AllChem.MMFFGetMoleculeProperties(moleculewH)

for atom in range(moleculewH.GetNumAtoms()):
    print(atom+1, "\t", properties.GetMMFFAtomType(atom), "\t", moleculewH.GetAtomWithIdx(atom).GetIsAromatic())

properties.SetMMFFVerbosity(2)
forcefield = AllChem.MMFFGetMoleculeForceField(moleculewH, properties)

Here below I attach an output screenshot printed from the above snippet of Python code. Only [atom | MMFF94 atom type | is atom aromatic] and bond stretching contributions are shown for sake of brevity.

I would expect to see right MMFF94 atom types for aromatic atoms and ring atoms to be recognized as aromatic.
I have the same problem e.g. for all nucleobases and also for thymine expressed in different ways such as CC1=CNC(=O)NC1=O or O=c1[nH]c(=O)[nH]cc1C

I'm using

RDKit version: 2022.03.4
OS: Ubuntu 20.04
Python version (if relevant): 3.9.1
Are you using conda? yes
If you are using conda, which channel did you install the rdkit from? pip
If you are not using conda: how did you install the RDKit?

Thank you in advance for any help!

[ptosco]

Ciao Andrea,

The atom types assigned by RDKit are correct, and so are partial charges; I have just verified the assignment with MOE and OpenBabel I get the same atom types and partial charges:

$ obenergy -ff MMFF94 thymine.mol

A T O M   T Y P E S

IDX     TYPE    RING
1       7       NO
2       3       AL
3       10      AL
4       3       AL
5       7       NO
6       10      AL
7       2       AL
8       2       AL
9       1       NO
10      28      NO
11      28      NO
12      5       NO
13      5       NO
14      5       NO
15      5       NO

Thymine is not aromatic according to the MMFF aromaticity model, which is why after assigning MMFF atom types the aromaticity flags are all set to False.

[apiserchia]

Dear Paolo,
thank you for your kind answer. I made a big mistake because I was thinking in "chemical terms", that is, since thymine is aromatic from the chemical point of view, then carbon atoms n.7 and n.8 should be aromatic carbons (MMFFTYPE: 37) instead of olefinic ones (MMFFTYPE: 2), as well as the other ring atoms having more "aromatic" atom types, but I didn't realize that MMFF aromaticity model excluded such molecules from being aromatic.
I understand this is due to specific MMFF aromaticity related rules that perceive benzene, pyridine, furan, etc... as aromatic molecules while nucleobases and other etherocyclic molecules as not aromatic.
As a last question I would like to ask if you have a(some) reference(s) where I may find specifically MMFF aromaticity model explained? In particular a reference where these MMFF criteria/rules are rationalized/listed in order to properly assign aromaticity according to MMFF aromaticity model without reverse engineering setMMFFAromaticity function. Unfortunately, as far as I read, I didn't find these rules listed in Halgren's MMFF94 papers.

Thank you again for your kind attention!