El Siglo XXI y el biocomercio en Colombia

Ideally, the parameters assigned to each molecule would be validated through comparison of the thermodynamic properties obtained from a molecular dynamics simulation with experimental data. However, finding experimental data to validate computational results is a difficult propo- sition. The GROMOS force field was validated using solvation free energy calculations,45 as was the ATB.9,11Such an undertaking could not be achieved here as experimental solvation free energy values tend to only be available for molecules with low molecular weights. NMR ex- periments are ubiquitous throughout chemistry, particularly within organic chemistry. An NMR spectrum provides a time averaged measure of the chemical shifts of NMR active elements within the molecule. Chemical shifts provide structural information. Thus, an NMR spectrum can be produced from a molecular dynamics simulation by calculating chemical shifts at each frame of the simulation, and averaging across all frames. As such, an alternative validation means, utilising NMR spectra was thus undertaken. Experimental NMR spectra were obtained from the Spectral Database for Organic Compunds (SDBS) which provides a large database of

8. Testing and Discussion

hydrogen and carbon NMR spectra,46and all molecules in CPT023 are present in the database. Method In order to calculate the NMR spectra, the CHARGE program was utilised,47which is a semi-empirical algorithm which determines1H NMR chemical shifts for a given structural geometry, taking into account a number of short and long range effects. Though CHARGE is capable of determining splitting due to coupling, calculations were performed without this capability enabled due to the reference data not providing coupling constants. Simulations were run in UA form, meaning that in order to calculate shifts correctly, the positions of non-polar hydrogen atoms needed to be generated. For the molecules simulated here, all the hydrogen atoms to generate were on tetrahedral carbons. Three types of carbon atoms required bonded hydrogen atoms to be placed: CH₁, CH₂ and, CH₃. In each case, the hydrogen atoms were placed on the points of a tetrahedron, centred on the central carbon atom and defined by the other substituents, assuming perfect tetrahedral geometry. In the case of CH₃, an initial hydrogen atom was placed randomly with respect to the fourth substituent on the central carbon in order to fix the orientation of the tetrahedron. Hydrogen atoms of the CH₁group where placed with a bond length of 108.83 pm, while the CH₂and CH₃hydrogen atoms were placed with a bond length of 109.45 pm. These bond lengths came from determining the mean of all the similar bond lengths for molecules within SRC9064, which had been QM optimised by the ATB as part of the parameter generation method. Further, a small amount of noise was added to each position, based on the standard deviation of the set of bond lengths, in order to simulate the vibration of the CH bond. To do so, a(r,θ,ψ)triplet was randomly generated. A value forrwas selected from a normal distribution withμ=0 andσ=2.399 pm for CH₁hydrogen atoms, and 1.601 pm for CH₂and CH₃hydrogen atoms;θ was selected from the uniform interval[0,π]; andψwas selected from the uniform interval[0,2π). The polar vector defined by this random triplet was then added to the previously calculated hydrogen atom position.

For every frame, 1_{H NMR shifts were calculated using CHARGE. For each atom, the mean}

shift was determined and plotted as a Lorentz distribution with width at half height of 0.5 Hz. The experimental spectra are given in figure 8.10 and the calculated spectra are given in fig- ures 8.11 to 8.17. Based on the shift values provided with the experimental spectra, reference spectra were also plotted using the same parameters.

Results Both the simulations with ATB and CherryPicker generated parameters tend to repli- cate the reference spectrum well. The majority of peaks in the 0 to 2 ppm range are due to hydrogen atoms on alkyl chains, with several bonds to the nearest hetero atom. The experimen- tal data had poor separation of these peaks, due to their very similar chemical environments. Large numbers of chemically similar hydrogen atoms were grouped into the same reported shift,

8.2. Parameterisation Tests

(a) (b)

(c) (d)

(e) (f)

(g)

Figure 8.10:Experimental1H NMR spectra for (a) molecule I, (b) molecule XXII, (c) molecule IX, (d) molecule VII, (e) molecule VIII, (f) molecule XVI and (g) molecule X. Spectra were obtained from the SDBS.46

8. Testing and Discussion

Figure 8.11: Calculated 1H NMR spectra of molecule I.

Figure 8.12: Calculated 1H NMR spectra of molecule XXII. The amine hydrogen shifts, la- belled with red arrows, are at 2.53 ppm for the ATB simulations, 1.66 ppm for the CherryPicker simulations and 3.39 ppm for the reference data.

8.2. Parameterisation Tests

Figure 8.13: Calculated 1H NMR spectra of molecule IX. The amine hydrogen shifts, labelled with red arrows, are at 1.12 ppm for both the ATB and CherryPicker simulations and at 1.76 ppm for the reference data.

Figure 8.14: Calculated 1H NMR spectra of molecule VII. The split hydrogen amide peaks, labelled with red arrows, are at 5.71 ppm and 5.33 ppm for the ATB simulations, and 5.68 ppm and 5.30 ppm for the CherryPicker. The single reference peak is at 5.60 ppm.

8. Testing and Discussion

Figure 8.15: Calculated 1H NMR spectra of molecule VIII. The split hydrogen amide peaks, labelled with red arrows, are at 5.71 ppm and 5.33 ppm for the ATB simulations, and 5.12 ppm and 4.78 ppm for the CherryPicker. The single reference peak is at 5.49 ppm.

Figure 8.16: Calculated 1H NMR spectra of molecule XVI.

8.2. Parameterisation Tests

Figure 8.17: Calculated 1_{H NMR spectra of} molecule X.

Figure 8.18:Time series of the H-N-C-O dihedral in the ATB parameter set (top) and CherryPicker parameter set (bottom) simulations of VII. They show fluctuation around 180°, indicating a lack of free rotation about the carbon–nitrogen bond.

8. Testing and Discussion

which when plotted results in sharper peaks in the reference spectrum than the corresponding experimental spectrum had.

Molecules VII and VIII are very similar amides, with the only difference being that the carbon chain is two carbon atoms longer for molecule VII. They also both show a splitting of the amide hydrogen atom shifts which is not apparent in the reference spectra. In both cases, the experimental spectrum, figure 8.10d and e respectively, shows a very broad, low peak, spanning as much as 1 ppm. As this is the case, the splitting may in fact be represented in the experimental spectrum. The cause of the splitting is due to the intra-molecular hydrogen bond between one of the amide hydrogen atoms and the amide oxygen atom. This can be seen in a time series analysis plot of the dihedral defined by the hydrogen atom that is not part of the hydrogen bond, the nitrogen atom of the amide, the carbon atom of the amide and the oxygen atom of the amide. Such a time series is given for the simulation of molecule VII using both the ATB and CherryPicker parameter sets in figure 8.18. The time series shows fluctuation around a dihedral angle of 180°, which indicates that free rotation about the carbon–nitrogen bond is hindered by the intra-molecular hydrogen bond.

Molecule XXII (figure 8.12) is a benzalkyl amine. Both ATB and CherryPicker give similar shifts for the hydrogen atoms of the aromatic ring system, with the CherryPicker results being slightly closer to the reference data. However, both perform poorly for the amine hydrogen atom shift, with large upfield movement relative to the reference data, 2.53 ppm for the ATB simu- lation and 1.66 ppm for the CherryPicker simulation compared with 3.39 ppm for the reference data. Though the experimental peak seen in figure 8.10b is fairly broad, it is not broad enough to account for such drastic differences. This potentially indicates an over shielding from the aromatic system in both cases. A similar issue arises with molecule IX (figure 8.13), though not to nearly the same extent. Here, the amine hydrogen atoms have a slight upfield shift, 1.12 ppm for both simulations, relative to the reference data of 1.76 ppm.