# Why we can't see 3-d structures of certain compounds on PubChem?

It's common to read statements like one given below on PubChem:

Conformer generation is disallowed since MMFF94s unsupported element.

What's the meaning of the statement, and why are certain compounds unsupported? And what is MMFF94s? This might be something really basic, but since I have never come across I asked it.

Most classical molecular force fields are parameterized for a set of elements and atom types. The MMFF94 method was designed for standard organic drug-like small molecules, so it has a limited set of elements (H, Li, C, N, O, F, Na, Mg, Si, P, S, Cl, K, Ca, Fe, Cu, Zn, Br, I) categorized into 99 atom types (e.g. mmffprop.par from the Open Babel implementation).

In the cases of the metals, they should be ions, not covalently bonded to any other element.

As in the case of many molecule force fields, there are sets of bond types, angle types, etc. as combinations of the atom types. Molecules that contain elements outside the parameterization, or combinations that aren't in the bond or atom parameters, are rejected.

As an example, the MMFF94 validation set is available: http://server.ccl.net/cca/data/MMFF94/MMFF94_dative.mol2.shtml

Beyond the limits of MMFF94 and MMFF94s themselves, PubChem3D had several limits indicated in the accompanying manuscript: Bolton et. al. "PubChem3D: a new resource for scientists" J Cheminf. (2011) v. 3, art. 32)

• Not too large (with ≤ 50 non-hydrogen atoms).
• Not too flexible (with ≤ 15 rotatable bonds).
• Consists of only supported elements (H, C, N, O, F, Si, P, S, Cl, Br, and I).
• Has only a single covalent unit (i.e., not a salt or a mixture).
• Contains only atom types recognized by the MMFF94s force field.
• Has fewer than six undefined atom or bond stereo centers.

In the case of molecules with undefined atom or bond stereo (e.g., E/Z) multiple stereoisomers were generated. Personally, I'd use these compounds with extreme care - it's not always obvious what the original PubChem entry represents.

Thus there are other records without 3D versions (e.g, they're large, have multiple covalent units, etc.)

• I'll add that the UFF method has rule-based parameters that cover most of the periodic table with various levels of accuracy, which would cover a few of these issues. (I'd personally love to see a newer version of UFF, e.g. with polarizable charges, but that's a research question.) Dec 29, 2019 at 16:52

Since it's been a couple of days, I guess I convert my comment into an answer.

MMFF94s is Merck Molecular Force Field. From what I understand, the issue is that this FF has certain limitations when it comes to electrostatic interaction, estimating polarizability and intramolecular interactions, which confuses the method trying to predict 3D conformers for ionic compounds, metal complexes or molecules with twisted geometry.

From the Concluding Discussion section of the original publication [1, p. 515]:

Despite encouraging success, certain limitations are evident. One of particular importance arises from the fact that MMFF94 uses static atom-centered charges. As such, it neglects both higher order multipoles and electrostatic effects that arise from molecular polarizability. Because of these simplifications, MMFF94, like a number of other force fields, employs “enhanced” charge distributions that emulate the effect of polarizability in amplifying electrostatic interactions for favorable contacts in a high-dielectric medium. Unfortunately, these enhanced charge distributions also amplify electrostatically unfavorable iteractions, whereas proper account of polarizability would diminish them. They also improperly enhance electrostatic interactions in gas-phase or low-dielectric environments. Furthermore, they may not be optimal for describing intramolecular interactions, and may thereby limit the ability of the force field to account for differences in conformational energies. Indeed, compounds containing two or more strongly polar functional groups in close proximity have proven to be the most problematic in this respect, though good results have been obtained in most cases to date. … Other significant limitations include: the overly simplistic nature of the bond-charge-increment scheme used to assemble the partial atomic charges …; the lack of conformational dependence of the resultant charges …; and the omission of bond-torsion (and certain other) cross terms needed to account for significant geometrical changes that can occur when a torsion angle varies,… an example being the elongation of an amide partial C−N double bond by up to 0.1 Å when conjugation is broken. A further significant limitation is that no account is taken of metal-ligand interactions beyond that afforded by a relatively simplistic model that includes only electrostatic and van der Waals nonbonded interactions.

### References

1. Halgren, T. A. Merck Molecular Force Field. I. Basis, Form, Scope, Parameterization, and Performance of MMFF94. Journal of Computational Chemistry 1996, 17 (5–6), 490–519. DOI: 10.1002/(SICI)1096-987X(199604)17:5/6<490::AID-JCC1>3.0.CO;2-P.
• Like I wanted to see the structure of Diboron Dioxide, shouldn't the 3-d structure of Diboron Dioxide $\ce{(BO)2}$ accessible? Dec 25, 2019 at 14:45
• @Zenix AFAIK, both $\ce{BO}$ and $\ce{B2O2}$ are unstable high-temperature molecular species existing only in gaseous phase somewhere above 1200 °C. It doesn't look like MMFF was designed for conditions like this, but I'm afraid I cannot give you a more precise answer. Dec 25, 2019 at 14:55
• @andselisk - your point is good that MMFF94 has significant, known limitations. It was, after all, designed in 1994. It's pretty good, although I'm surprised there hasn't been much effort for a replacement over the last 25 years! Dec 25, 2019 at 19:07