A case study of spiroreticulatine
For the efficiency and accuracy of ‘Dooerafa’, spiroreticulatine, a previously reported new alkaloid which was isolated from the marine sponge Fascaplysinopsis reticulata and unambiguously determined by X-ray diffraction , was used as a verification module.
Spiroreticulatine displayed good 1H and 13C NMR experimental data while HMBC cross peaks can only establish fragments of the structure. And there were at least two structures with good relation with 2D NMR experiments (Fig. 1). The problem of structure conformation can be easily distinguished by DFT-GIAO calculations of 13C NMR data for now. But the real problem could be which kind and how many of the structural candidates the compound implied. So the ‘Dooerafa’ for spiroreticulatine was performed.
Because 2J and 3J correlations are ambiguous, which led the structure assignment difficult and more uncertain, only the segments which were determined unambiguously by the very certain 1H-1H COSY correlations and the HMBC correlations of methyl groups were assigned as the original issues (meta-groups, Figs. 1 and 2A). In that case, there are six meta-groups A-F for spiroreticulatine (Fig. 2A). However, there are 16 connection points (consistent with Additional file 1 Section 2) for shaping up a whole molecule which took about two weeks long for all the structure assembling in the preliminary ‘Dooerafa’.
For the efficiency, 11 meta-structures (eleven assembling types for meta-groups A, C-F on B) which were lowered down to 14 connection points were constructed based on the grafting method and were used for the assembling calculation. And that took less than 16 hours (Fig. 2B). Then 854 group data by Python programming code based on the deduced connection points (see Additional file 3) were obtained and were converted to chemical structures.
Subsequently, five kinds of energies related to stability of the structures including stretch, bend, stretch-bend, torsion and total energy were calculated automatically by Python (see Additional file 3) in Chem 3D software using MM2 molecular mechanic force field [23, 24], because there must be a chemical reasonability for the natural products under a certain structural assembling strategy. That is, the natural products with more stable chemical features tend to be the real and especially chemically comfortable structures. As a result, 50 structures showing lower energy before the inflection point in the most important trend of the total energy were selected to calculate 1H and 13C NMR data using DFT-GIAO method (Fig. 3).
Furthermore, a related analysis of the fitting between experimental and calculated data using coefficient of determination (R2) as well as mean absolute error (MAE) and truncated absolute difference (TAD) gave us the correct structure of spiroreticulatine which is the 5th one in all the 854 structure candidates. Other structures displayed good but not best match of 13C NMR data and were contradicted with the HMBC correlations. Thus, the verification of spiroreticulatine indicated that ‘Dooerafa’ is suitable for determining the structure of natural products.
Then, the ‘Dooerafa’ was applied to confirm the structure of an unknown alkaloid named aaptourinamine which was lack of the 2D NMR correlations as well. Firstly, the structure of aaptourinamine was preliminarily analysed by NMR data.
Structural investigation of aaptourinamine
Aaptourinamine was isolated from marine sponge Aaptos suberitoides. It has a molecular of C12H7ON3 as deduced from HRESIMS data m/z 210.0661 [M + H]+ (calcd for C12H8ON3, 210.0662) and 232.0481 [M + Na]+ (calcd for C12H7ON3Na, 232.0481) supported by the 1H and 13C NMR spectral data (see Additional file 1 Section 6). It was a potentially new compound with good purity and distinct 1H and 13C NMR experimental data having no match with the 13C NMR database in MICRONMR . However, it had high C/H ratio and was lack of efficient HMBC cross peaks to be unambiguously determined, neither is it possible for efficient 15N NMR due to the insufficient sample amount .
With the aid of HMQC and 1H-1H COSY experiments, aaptourinamine displayed one NH proton at δH12.31 (1H, s), and six olefinic protons at δH8.71 (2H, one doublet J = 6.85 Hz and one overlapped singlet), 8.03 (1H, d, J = 6.85 Hz), 7.29 (1H, t, J = 6.82 Hz), 7.25 (1H, t, J = 2.57 Hz), 6.95 (1H, dd, J = 2.35, 2.05 Hz) in 1H NMR. The profile of an aromatic compound was shown in 13C NMR (DEPT) displaying eleven olefinic carbons (five olefinic CH) with chemical shifts around δC105-134, as well as one carbonyl group at δC168.8. The 1H-1H COSY and HMBC correlations indicated a dihydropyrrole moiety in which the crucial cross peaks of δH7.25/δC122.4 and δH8.03/δC122.4 were distinguished from a cross peak group of all the olefinic protons with the adjacent δC122.4/122.5. Besides, there must be an imine group except for five pairs of double bonds and the carbonyl group according to HRESIMS data. Thus, there are three potential core structure candidates A-C except for D with an unreasonable location of the carbonyl group (Fig. 4A), because chemical shift of the carbonyl carbon in D indicating macrocyclic structure candidates will be greatly deshielded compared with δC168.8 .
To save calculation time, the grafting method (see Additional file 1 Section 2) based on the core structure candidates was applied to obtain the meta-structures with 14 connection points (see Additional file 1 Section 2) as in the case of candidate A (Fig. 4B). Thus, there are totally 44 meta-structures as shown in Figs. 4 and 5, which was constructed on the basis of that an imine group and a carbonyl group were certain and were connected with the core structures A-C or outside the core structures to make sure of less than 14 connection points totally for each meta-structure.
Subsequently, 2507 structures were drawn manually by the group data obtained through the program code in Python. These structures showed very huge diversity with varied kinds of assembling. Then, the energies related molecular stability were calculated by Python controlling the Microsoft Excel and Chem 3D. The molecular energy in mechanic force field displayed an interesting trend for the structural change. And there is an inflection point at which the structures tend to be unstable with a sudden changing especially of the torsion, stretch-bend and total energies (Fig. 6, see Additional file 2), which indicated the structures are chemically unfavourable in the case that the main structural moieties of meta-groups have been confirmed based on the crucial experimental NMR data. So there must be a very knowledgeable reason to shape up a complex and chemically unfair structures existing in nature .
Then 150 structures before the inflection point showing lower energy were calculated for 1H and 13C NMR using DFT-GIAO method (Fig. 6, see Additional file 1 Section 4) and analysed the linear regression parameters which were used to reflect the fitting between experimental and calculated data. Finally, the correct structure of aaptourinamine (A-205, the 9th one in the 2507 candidate structures, see Additional file 1 Section 5) with best match of calculated NMR data was shown as imidazo[4,5,1-ij]pyrrolo[3,2-f]quinolin-7(8H)-one which is structurally related to aaptamine alkaloid as predicted in the biosynthesis pathway in Scheme 1.
Therefore, aaptourinamine is shown as a new scaffold of aaptamine family which has totally different type featured by formation of a pyrrolo[3,2-f]quinolone core which is firstly found in nature, indicating an intriguing profile of the biosynthesis and bioactivity. The chemical shift of the carbonyl carbon C-8 at δC168.8 (Table S2, see Additional file 1) in aaptourinamine is greatly upshielded comparing to the normal ketone group probably due to the large and favourable conjugated system as in the case of exiguamine A [27b]. And the nitrogen atoms must play an important role for the unique conjugated aromatic system of 18 π-electrons over 16 centers .
To summarize, a calculation study of aaptourinamine by a time-saving module-assembly based calculation was performed in ‘Dooerafa’ which was validated through spiroreticulatine with structure determined by X-ray diffraction, providing the credible and accurate result. In this method, first, meta-groups are very important which were assigned from the believable NMR and HRESIMS experimental data. Then the grafting method is practical and efficient to obtain the meta-structures to save calculation time. Subsequently, the self-created program code in Python which shows all group data converting to the structure candidates is the first to report for determining the structure. And running the code only takes less than 16 hours, indicating the efficiency of ‘Dooerafa’.
On the other hand, DFT-GIAO calculation of NMR data has been proved to be a reliable method to predict the right structure. Especially 13C NMR data which is almost not changing in different solvent in the practical experiments is absolutely suitable for the DFT calculation . Nevertheless, 1H NMR data is impressible to the solvent and is not a good choice to determine structure as in the cases of spiroreticulatine and aaptourinamine in the present study. Thus, a logical use of mechanic force field and quantum chemical theory of ‘ring-contraction strategy’ is the key point for the study.
Hence, the molecular energies trend in mechanic force field which showed clear inflection points is suitable to select the stable structures with great efficiency (Fig. 7), in which the correct structure of aaptourinamine occupied 9th in the top 150 structures. And the molecular energy with DFT calculation generally displayed a similar trend but showed a closer relation with the correct structure. Nevertheless, the 150 structures with lower energies are biogenically irregular but chemically favourable. Finally the right skeleton of the natural product was figured out according to the NMR calculation showing the best coefficient of determination (R2). Thus, the structures which were ticked from the numerous candidates and showed the best match with the experimental data must be the right and unique natural occurring structure.