Wheldone Structure Elucidation

March 24, 2025

by Mikhail Elyashberg, Leading Researcher, ACD/Labs

---

Computer Assisted Structure Elucidation of Wheldone Using Structure Elucidator Suite

Oberlies and co-workers [1] isolated a new natural product, Wheldone (A) as a result of investigation of a coculture of Aspergillus fischeri (NRRL 181) and Xylaria flabelliformis (G536). Its structure was elucidated by one- and two-dimensional NMR and mass spectrometry. Compound A displayed cytotoxic activity against breast, ovarian, and melanoma cancer cell lines.

A

We used data from this article to challenge Structure Elucidator Suite. Compound A was obtained as a white amorphous powder with a molecular formula of C₂₅H₃₄O₆, as determined via HR-ESI-MS along with ¹H, ¹³C, edited-HSQC and HMBC NMR data. The NMR data used are shown in Table 1.

Table 1. NMR spectroscopic data of wheldone [1].

The data were entered into the program which created a Molecular Connectivity Diagram (MCD) shown in Figure 1.

Figure 1. The Molecular Connectivity Diagram (MCD) of wheldone . Hybridizations of carbon atoms are marked by the corresponding colors: sp² – violet, sp³ – blue. Labels “ob” and “fb” are set by the program to carbon atoms for which neighboring with heteroatom is either obligatory (ob) or forbidden (fb). The HMBC connectivities are marked by green arrows. The evident C=O bond for the atom with chemical shift of 208.7 was drawn manually.

Checking the MCD for the presence of contradictions in HMBC data was completed and the following program message was shown: The minimum number of non-standard connectivities is 1. This means that the HMBC data contain at least one correlation whose length exceeds three chemical bonds. In this case, Fuzzy Structure Generation (FSG) should be used [2]. FSG combined with ¹³C chemical shit calculation was run with the options that were selected automatically by the program. Structure generation was completed in 5 s with the following results: k = 97 → (structure filtering) → 36 → (duplicate removal) → 36.

The ¹³C chemical shifts were predicted for all the structures generated using the three algorithms implemented into Structure Elucidator (HOSE code-based, neural network and incremental approach), and structures were ranked in increasing order of average deviations of calculated chemical shifts from experimental ones. The six top ranked structures are shown in Figure 2.

Figure 2. The six top ranked structures of the output file. ¹³C chemical shift prediction was carried out using the HOSE code based method, the neural networks, and the incremental approach. Average deviations of ¹³C chemical shifts determined by these methods are denoted as d_A, d_N and d_I correspondingly. Each atom is colored to mark a difference between its experimental and calculated ¹³C chemical shifts. The green color represents a difference between 0 to 3 ppm, yellow was >3 to 15 ppm, red > 15 ppm. The red arrow shows a nonstandard HMBC correlation detected by the program during the fuzzy structure generation.

The best structure, #1, is characterized by minimal average deviations calculated by all three methods. However, the HOSE codes predicted chemical shift for atom 5 differs from the experimental value by 20 ppm (the atom is colored red). At the same time, the predicted values of chemical shifts of the remaining atoms are very close to the experimental ones. The HOSE protocol for calculating the chemical shift of atom 5 showed that the database did not contain sufficient structures on the basis of which the chemical shift of atomic 5 could be calculated with greater accuracy; the structures found in the database were only similar up to 2 spheres (shells) away, so the prediction was rather uncertain. Such cases are quite rare, but they are likely in principle, since the number of possible combinations of atoms in organic molecules is almost infinite. At the same time, the values of deviations provided by the incremental and neural network methods in this case (2-2.5 ppm) are typical of a correct structure.

To evaluate the correctness of structure #1, the DP4 probabilities were calculated for all six structures shown in Figure 2.

Figure 3. Correct structure probabilities. DP4_A, DP4_N and DP4_I are probabilities of structure correctness calculated by the program.

Figure 3 shows that the probability of structure #1 correctness is 100%. However, the presence of a “red” atom in this structure makes it necessary to confirm it with additional experiments and/or DFT calculations of chemical shifts.

 

Interestingly, structure A proposed by the authors [1] was not generated. Figure 4 shows structures A and #1. We see that structure A contains 6 non-standard HMBC correlations, while structure #1 has only one correlation of this type. These non-standard correlations were allowed by the authors [1] when deriving structure A. Obviously, structure A could not have been generated, since HMBC correlations have a length of 2-3 chemical bonds by default in Structure Elucidator [2].

A

#1

Figure 4. Nonstandard HMBC correlations in structures A and #1.

We entered structure A in the program and calculated the ¹³C chemical shifts. Figure 5 convincingly shows that this structure is erroneous, which follows from the very large deviations calculated by all methods. Obviously, if such calculations were made by the authors, they would see that their structural hypothesis is not correct.

Figure 5. Proposed structure A. The results of ¹³C chemical shift prediction are shown.

The correctness of structure #1 was recently confirmed by the same group [3] which carried out revision of structure A. This time, structure #1 was elucidated using HMBC and 1,1-HD-ADEQUATE data and it was confirmed using DFT calculation of NMR chemical shifts and crystallographic analysis.

The revised structure of wheldone with ¹³C chemical shift assignments is shown below:

The example described shows how important verification of a structural hypothesis is with any of the widely available fast empirical methods for NMR chemical shifts prediction. In most cases, verification prevents deducing and publication of erroneous structures.

References

1. Knowles, S. L.; Raja, H. A.; Isawi, I. H.; Flores-Bocanegra, L.; Reggio, P. H.; Pearce, C. J.; Burdette, J. E.; Rokas, A.; Oberlies, N. H. Wheldone: Characterization of a Unique Scaffold from the Coculture of Aspergillus fischeri and Xylaria flabelliformis. Org. Lett. 2020, 22, 1878−1882.
2. M.E. Elyashberg, A.J. Williams. Computer-based Structure Elucidation from Spectral Data. The Art of Solving Problems. Springer, Heidelberg, 2015, 454 p. http://www.springer.com/978-3-662-46401-4
3. M. Rangel-Grimaldo, C. E. Earp, H. A. Raja, J. S. Wood, L. Mardiana, K. L. Ho, A. Longcake, R. T. Williamson, L. Palatinus, M. J. Hall, M. R. Probert, N. H. Oberlies. Wheldone Revisited: Structure Revision Via DFT-GIAO Chemical Shift Calculations, 1,1-HD-ADEQUATE NMR Spectroscopy, and X‑ray Crystallography Studies. J. Nat. Prod. 2024, 87, 8, 2095–2100

---