Elucidating the Mechanism and Predicting the Selectivity of the bis(Pyridine) Silver Permanganate Oxidatiin

The Movassaghi group has recently reported several total syntheses of epipolythiodiketopiperazine derivatives (ETPs), a class of natural products with potent biological activity. A crucial step in their synthesis oxidizes two to four tertiary C-H bonds to form tertiary alcohols using (bis)pyridine silver permanganate, but this essential synthetic step is not well understood and is not always successful. Because of the utility of this reaction in ETP synthesis as well as its potential to be useful in other systems, it was selected as an important reaction for study. Toward this end, a library of simple molecules with single potential sites of oxidation, hydantoins, were synthesized, and relative rate data was obtained via competitive rate experiments under the oxidation conditions. Using the rate data, a correlation was generated, comparing computed descriptors to the relative oxidation rates. The model generated for the hydantoins was able to predict whether 13 of 16 ETP precursors were able to undergo oxidation in this system, showing large energy barriers for ETP precursors which weren't oxidized and small energy barriers for those that were oxidized. Additionally, a single parameter * was identified, which also produced an accurate prediction for 13 of the 16 ETP precursors, with those undergoing oxidation below a certain threshold and those not undergoing oxidation above that threshold. In the context of the Movassaghi group's oxidation, this thesis describes a tactic by which simple molecules can be used to predict reaction outcomes in more complex systems.

Elucidating the Mechanism and Predicting the Selectivity of the bis(Pyridine) Silver Permanganate Oxidation by Amanda Bischoff A Senior Honors Thesis Submitted to the Faculty of The University of Utah In Partial Fulfillment of the Requirements for the Honors Degree in Bachelor of Science In The Department of Chemistry Approved: ______________________________ Dr. Matthew Sigman Thesis Faculty Supervisor _____________________________ Dr. Cynthia Burrows Chair, Department of Chemistry _______________________________ Dr. Thomas Richmond Honors Faculty Advisor _____________________________ Dr. Sylvia D. Torti Dean, Honors College December 2016 Copyright © 2016 All Rights Reserved ABSTRACT The Movassaghi group has recently reported several total syntheses of epipolythiodiketopiperazine derivatives (ETPs), a class of natural products with potent biological activity. A crucial step in their synthesis oxidizes two to four tertiary C–H bonds to form tertiary alcohols using (bis)pyridine silver permanganate, but this essential synthetic step is not well understood and is not always successful. Because of the utility of this reaction in ETP synthesis as well as its potential to be useful in other systems, it was selected as an important reaction for study. Toward this end, a library of simple molecules with single potential sites of oxidation, hydantoins, were synthesized, and relative rate data was obtained via competitive rate experiments under the oxidation conditions. Using the rate data, a correlation was generated, comparing computed descriptors to the relative oxidation rates. The model generated for the hydantoins was able to predict whether 13 of 16 ETP precursors were able to undergo oxidation in this system, showing large energy barriers for ETP precursors which weren’t oxidized and small energy barriers for those that were oxidized. Additionally, a single parameter * was identified, which also produced an accurate prediction for 13 of the 16 ETP precursors, with those undergoing oxidation below a certain threshold and those not undergoing oxidation above that threshold. In the context of the Movassaghi group’s oxidation, this thesis describes a tactic by which simple molecules can be used to predict reaction outcomes in more complex systems. ii TABLE OF CONTENTS ABSTRACT ii ACKNOWLEDGMENTS iv INTRODUCTION A. HAMMETT ANALYSIS 1 B. MODELING IN MODERN ORGANIC CHEMISTRY 3 C. EPIPOLYTHIODIKETOPIPERAZINE SYNTHESES 5 D. BIS(PYRIDINE) SILVER PERMANGANATE OXIDATION 5 METHODS A. KINETIC MEASUREMENT CONDITIONS 7 B. OBTAINING COMPUTATIONAL DATA 9 C. MODELING 10 RESULTS A. NUMBERING SCHEMES 12 B. SOLUBILITY DATA 13 C. RATE DATA 16 D. COMPUTATIONAL DATA 18 E. MODELING DATA 21 DISCUSSION 24 REFERENCES 32 iii iv ACKNOWLEDGMENTS First and foremost, I would like to thank my undergraduate research advisor Matthew Sigman. His advice and patience as I’ve faced challenges in this project have been invaluable, as has his guidance in determining my plans to go to graduate school and pursue chemistry further. He had confidence in my project and my abilities even when I didn’t. I would also like to express my appreciation for my collaborators, Mohammad Movassaghi and Brandon Nelson. Their idea motivating the project and advice have been essential to its success. In the Sigman lab, Zach Niemeyer has also been instrumental in the project’s completion, and I really appreciate the many times I’ve been able to bounce ideas off of him and for the key insights he provided which drove the project forward. I would like to thank the entire Sigman lab for their help with my project and for fostering an environment where I felt free to ask questions and seek help when I needed it. In particular, I would like to thank Iris Guo for use of her Matlab script for modeling and Margaret Hilton for answering my questions and extensive help with the draft for my manuscript. Most importantly, I would like to thank my family for their support to me throughout college. My dad has always told me I can do anything I put my mind to and encouraged me no matter how difficult things were, and my stepmom has also been encouraging and unfailingly supportive. Lastly, I would like to thank my little sister, my best friend and my confidant, for the many times she listened to me complain and her outside perspective on my work. 1 INTRODUCTION A. HAMMETT ANALYSIS There is a common saying in the physical sciences that “an hour in the library is worth a day in the lab.” Sometimes, taking a closer look into the literature helps foster new ideas or build on a finding whose significance was originally missed. Although linear free energy relationships in organic chemistry have been around for over 80 years (1), their significance is still being realized. Linear free energy relationships (LFERs) describe how a measurable characteristic of a molecule is related to the activation energy of that molecule in a particular reaction. One of the seminal descriptions of an LFER examined how different substituents on the phenyl ring of benzoic acid influenced its acidity (Fig 1) A value x, called a Hammett parameter, was assigned to each substituent by comparing it to benzoic acid using Equation 1, in which KH is the reaction rate constant for benzoic acid and K X is the reaction rate constant for substituted benzoic acid x. 𝐾 𝜎𝑋 = log (𝐾𝐻 ) 𝑋 (1) Figure 1. General Reaction for Generation of Hammett Parameters The parameter x was found for both meta and para substituents, resulting in a scale that defines electron density. Correlating the x parameters for different substituents to the rate of other reactions for those substituents is called a Hammett plot. These Hammett 2 plots can thus distinguish whether increasing electron density increases or decreases the rate of reaction, providing powerful information about the reaction mechanism (1). While Hammett parameters describe resonance and inductive effects, they provide no information about the steric and polar effects of different substituents. To probe these effects, a new analysis was performed in which the rate of methyl ester hydrolysis was measured with varied hydrocarbon chains (Fig 2). In this manner, a steric parameter Es was defined using Equation 2, in which ks is the reaction rate constant for methyl ester s, and kCH3 is the rate constant for methyl acetate, both under acid catalysis. 𝐸𝑠 = log (𝑘 𝑘𝑠 𝐶𝐻3 ) (2) Figure 2. General Reaction for Generation of Taft Parameters. By contrasting the acid- and base-catalyzed pathways of the ester hydrolysis, a parameter *was also defined to describe polar effects. This parameter increases with enhanced polarizability and is described by Equation 3, in which B stands for base-catalyzed and A stands for acid-catalyzed. The coefficient is employed to make the Taft parameters comparable in magnitude to the Hammett parameters. 1 𝜎 ∗ = 2.48 [log (𝑘 𝑘𝑠 𝐶𝐻3 ) − log (𝑘 𝐵 𝑘𝑠 𝐶𝐻3 ) ] (3) 𝐴 Like the Hammett parameters, relating the Taft parameters to the effects of different substituents on reaction rate yields mechanistic information by describing how steric and polar effects influence the reaction rate (2). While Hammett and Taft parameters are informative, these parameters are specific and cannot always be correlated well with reaction outcomes for unique systems. 3 However, extensive tables of Hammett and Taft parameters for different substituents are available, and even correlating reaction rates from a few substrates to these parameters can reveal interesting mechanistic information. B. MODELING IN MODERN ORGANIC CHEMISTRY Today, much of biological, physical, and theoretical chemistry is greatly informed by computation. In contrast, many current research advances in organic chemistry are based on empirical evidence. Those who have studied organic chemistry thoroughly can gain a qualitative, intuitive feel for what conditions may lead to an unforeseen product which informs their discovery. In practice, however, many of the new discoveries in organic chemistry are made through extensive screening of substrates, catalysts, ligands, solvents, temperatures, and more. Combining all of these variables, it quickly becomes apparent that developing one new reaction expends a lot of time and resources without a lot of intellectual gain. The field could thus greatly benefit from a method by which reaction mechanisms could be analyzed quantitatively and optimized computationally (3). With the advent of powerful computational tools, Hammett and Taft parameters, as well as LFERs in general, become much more versatile than they were when discovered. Hammett plots can be generated easily in simple programs like Microsoft Excel. Even more powerfully, programs like Gaussian can easily generate energyoptimized structures and extract Hammett values as well as a vast range of other values for structures, which haven’t been studied experimentally. With just a bit of a learning curve, multivariate analysis tools like Matlab allow researchers to fit their research to multiple parameters at the same time. Although these are relatively simple calculations to perform, many organic chemists do not make use of them. 4 The Sigman lab at The University of Utah has made strides toward using computation to inform reaction development in recent years (4). In order to accomplish this, a training set is first selected with electronic and steric variations at several positions. For this training set, several parameters are computationally derived, which describe variations in the training set, and a linear regression analysis is performed to correlate the computed parameters to experimental results (such as enantiomeric excess or rate constant). This correlation can then be used to predict how other reaction components may perform under the reaction conditions. The Sigman group has been able to use this method to computationally design catalysts which will produce a higher enantiomeric excess (5) or predict the selectivity of a reaction for unique substrates (6). This type of prediction has been used by the Sigman group as well as the, Toste (7), Doyle (8), Sanford, and Minteer (9) groups among others (10) to tailor their reaction designs, proving the utility of modeling in a wide range of systems. Ultimately, this allows higher-performing ligands to be designed and reduces the amount of time and effort. Because the field of computational design to inform organic chemistry is relatively new, there are many challenges as well as opportunities for growth. Several questions remain to be addressed, including how to streamline the computational process and make it accessible to organic chemists who don’t have a computational background. Additionally, it remains unclear how extensively computation can be used and whether information gained by studying one reaction system could be used to inform another system entirely. My thesis project aims to address this second question by examining a 5 model system and finding approaches to apply the computational findings from the model system to a unique, and more complex, situation. C. EPIPOLYTHIODIKETOPIPERAZINE SYNTHESES Epipolythiodikeopiperazines (ETPs) are a class of molecules with promising biological properties, including antibacterial (11), anticancer (12), antiviral (13), antifungal (14), immune-related (15), and other biological activity (16). Until recently, however, there was no reported route for the synthesis of these molecules. Over the past several years, Movassaghi and coworkers developed a synthesic strategy for ETPs, which has allowed for the synthesis of several analogs for biological testing (17) (Figure 3). Due to the complex scaffolds and promising uses of these molecules, their syntheses were identified as good candidates for study using computational modeling. Figure 3. ETPs synthesized by Movassaghi and coworkers. D. BIS(PYRIDINE) SILVER PERMANGANATE OXIDATION One of the key steps of Movassaghi and coworkers’ ETP syntheses is a late-stage, bis(pyridine) silver permanganate-mediated tetrahydroxylation (Figure 4). While this reaction has been effective in several syntheses, it has been ineffective for a variety of 6 other substrates for reasons not readily understandable. Thus, time and effort has been expended on precursors, which fail at a later synthetic step. Additionally, this reaction is attractive to study because it has been shown to provide a method by which to selectively oxidize tertiary C–H bonds α to amides to form quaternary centers, which could provide utility in many syntheses if it were better understood. Me HS N Me O H H N O Me HO N O SH HS O N N N H H O Me SH Me Me H H N N N HO O O OH N N N H H O Me OH Me Figure 4. Oxidation of tertiary C–H bonds in ETP synthesis. Motivated by the importance and unpredictability of this reaction, the Movassaghi group sought to collaborate with the Sigman lab on a mechanistic and computational analysis to determine which factors influence a substrate to or not to undergo oxidation. This proposal was challenging because the oxidation of the ETP precursors contains two non-independent oxidations within the same molecule. Because the substrates used in the oxidation require six to eight steps to synthesize, it was an attractive idea to use less complex molecules and attempt to extrapolate the results to the more complex scaffolds of the ETP precursors. In doing so, we could understand the mechanism of the oxidation better, use the mechanistic understanding to predict reaction outcomes, and show that complex systems can be successfully modeled using simpler substrates. 7 METHODS A. KINETIC MEASUREMENT CONDITIONS My collaborators Brandon Nelson and Zach Niemeyer initiated this project by identifying a small molecule, which was easily synthesized and could approximate the larger ETP precursor motifs. Together, they identified a class of substrates, hydantoins, which could be modified at the C5 and N1 positions to yield a library of substrates with electronic and steric variation at these sites (Figure 5). These molecules also had the benefit of containing a single sight with the potential for oxidation. Brandon developed the synthetic routes to obtain these substrates and synthesized them for me to use in relative rate studies to determine their oxidation. Figure 5. Hydantoins synthesized and their variations. Initially, it was hypothesized that the rate experiments could be performed on individual substrates and monitored using in-situ IR. We theorized that we could follow product formation by monitoring the appearance of an alcohol peak as the reaction proceeded, but the instrument used did not encompass the alcohol range. While other peaks were identified for potential use in monitoring, these were often difficult to distinguish from irrelevant peaks and provided irreproducible results. Additionally, the reaction proceeded too quickly for some substrates for the instrument to record, so initial rate measurements couldn’t be obtained (the in-situ IR can only take time points ever 15 seconds). Due to these challenges, several other approaches were attempted. 8 Monitoring the reaction was also attempted by taking aliquots, quenching, and monitoring by GC, but it again proved problematic to run these reactions due to how rapidly they proceeded and irreproducibility. Attempts to use in-situ NMR also failed, with difficulty obtaining spectra due to the radical nature of the oxidant. We theorized that the problematic element in the reactions may have been the oxidant’s solubility, which is used at concentrations of 2 mM in the Movassaghi group’s syntheses. Thus, solubility was measured by UV. This was done by weighing out a certain amount of the oxidant py2AgMnO4, performing serial dilutions of this solution, and measuring the UV absorbance for each dilution. By monitoring when concentration no longer correlated with absorbance, the threshold of solubility could be obtained. This was performed for the oxidant py2AgMnO4 as well as another oxidant which had been used in one of the Movassaghi group’s syntheses as an alternative, nBu4NMnO4 as we considered doing further experiments comparing the two oxidants. These measurements were performed at 980 nm for Py2AgMnO4 and at 900 nm for nBu4NMnO4 by subtracting the absorbance of a blank cuvette from the absorbance of the oxidant at various concentrations. Acetone and acetonitrile were measured to probe which solvent would give the best rate data. Due to limited solubility of Py2AgMnO4 in both acetone and acetonitrile, it was determined that relative rate measurements should be made instead of individual rate measurements. Additionally, the oxidant was most soluble in acetonitrile, so this solvent was used for the rate experiments. By doing this, the solubility of the oxidant would be less of a factor because it was used at lower concentrations. Thus, the following general procedure was used for obtaining relative rates: 9 Figure 6. General Procedure for obtaining relative rate measurements. Into three 5-mL vials was weighed bis(pyridine)silver(I) permanganate (4.00 equiv, 0.0400 mmol, 15.4 mg). The vials were then equipped with stir bars and purged with nitrogen for 10 minutes, and 0.5 mL acetonitrile was added to each. In a separate vial, hydantoin X (3.00 equiv, 0.300 mmol, 57.0 mg) was dissolved in acetonitrile (0.75 mL), and in a separate vial, the hydantoin substrate X (3.00 eq, 0.300 mmol) and 1methylnaphthalene (3.00 equiv, 0.300 mmol, 42.6 L) were also dissolved in acetonitrile (0.75 mL). The two solutions were combined, and a portion (0.5 mL) of the resulting solution was transferred via syringe to each reaction vial under a nitrogen atmosphere. After 1 h, each reaction mixture passed through a celite plug and examined using gas chromotography. Thus, the relative rate data was obtained and converted to ∆∆G‡ according to Equation 4, in which r1 and p1 are the reactant and product of the reference substrate-1 (hydantoin 19a), and ri and pi are the reactant and product of the substrate X. 𝑝 ∗𝑟 ∆∆𝐺 ‡ = 0.001986 ∗ 298 ∗ ln (𝑝𝑖 ∗𝑟1 ) 1 𝑖 (4) B. OBTAINING COMPUTATIONAL DATA Gaussian 09 software (18) was used to obtain DFT calculations. The hydantoin and diketopiperazine geometries were optimized at the M06-2x/def2tzvp level of theory with an ultrafine integration grid. The MO6-2x functional was used for its utility in structure optimization for kinetics and non-covalent interactions (19). The def2tzvp basis set was used because it quantitatively produces very slight 10 errors (1 pm in bond length or 1° in bond angle) but requires much less computational memory than more rigorous basis sets (20). Several infrared vibrations and intensities, 13C and 1H NMR shifts, NBO charges, and various molecular angles and distances were obtained from these optimized structures in Gaussview. These structures were also used in Molecular Modeling Pro® (21) to obtain sterimol, Hammett, and Taft parameters and the steric hindrance parameter. Several relevant parameters for the hydantoins and diketopiperazines are shown in the Tables 1-4. Parameters highlighted in red were used for the models. C. MODELING The initial plan for modeling the data was that a training set would be used to obtain an initial model, and a validation set would be used to confirm the validity of that model. The original training set consisted of the hydantoins 19a-27a shown in Figure 8. A model was obtained for these initial results based on several parameters, and then a validation set was selected which had variations in the important parameters from the initial model. Based on predictions from the initial model, I selected a validation set which was synthesized by my collaborator, Brandon Nelson (hydantoins 28a-34a). However, this validation set revealed that the original model did not account for all variations. Thus, to gain a more comprehensive model, hydantoins 35a-39a were synthesized, and all substrates were used for modeling. The sole substrate used for validation was an analogue of 20a with deuterium in place of hydrogen in the oxidized position (45a). This was done in order to discern whether the model could predict a kinetic isotope effect. 11 In order to correlate the calculated parameters with the measured rate data, a multivariate linear regression analysis was performed using Matlab® R2015b software (22). Using a script written by Iris Guo, several models with adequate R 2 values using one to seven of these parameters to predict the relative transition state energies ∆∆G ‡ determined by relative rate experiments were generated. Correlations with a high R2 indicate that the model is a good approximation of experiment. A leave-one-out analysis was also employed to determine the predictive ability of the models. The model selected for further study was the one with the best Q 2 value as well as parameters which were easiest to interpret in terms of the mechanism of the oxidation. (23) Once a sufficient model had been obtained, the parameters composing the model were examined individually for different subsets of the hydantoin substrates in which either R1 or R2 did not vary. The correlations from these analyses were used to describe how different substrate characteristics affected their rate of reaction. The computationally derived model was also used to determine ∆∆G‡ values for ETP precursors, which had previously been submitted to the oxidation conditions by the Movassaghi group. While there was no relative rate data available for these substrates, several of them had undergone oxidation of both tertiary hydrogen positions, while others had only undergone a single or no oxidation event. The goal of finding the ∆∆G ‡ for these substrates was to determine whether there was a threshold over which the ETP precursors did not react. Toward this end, the computational parameters were also examined individually to see whether any one of them had an “on-off” threshold where the parameters for sites that were not successfully oxidized were all higher (or lower) than sites that were successfully oxidized. 12 RESULTS A. NUMBERING SCHEMES Figure 7. a) Numbering scheme for hydantoin molecules. b) Numbering scheme for hydantoins which aren’t oxidized under the reaction conditions and all diketopiperazines which are or are not oxidized under the reaction conditions. 13 B. SOLUBILITY DATA Py2AgMnO4 in acetone 0.6 Absorbance 0.5 0.4 0.3 0.2 0.1 0 0 0.02 0.04 0.06 0.08 Oxidant concentration (mM) 0.1 0.12 0.1 0.12 Py2AgMnO4 in acetonitrile 0.4 0.35 Absorbance 0.3 0.25 0.2 0.15 0.1 0.05 0 0 0.02 0.04 0.06 0.08 Oxidant concentration (mM) Fig 8. Solubility of Py2AgMnO4 in acetone and acetonitrile as determined by UV spectroscopy. 14 (nBu)4NMnO4 in acetone 0.3 0.25 Absorbance 0.2 0.15 0.1 0.05 0 0 0.02 -0.05 0.04 0.06 0.08 0.1 0.12 0.1 0.12 Oxidant concentration (mM) (nBu)4NMnO4 in acetonitrile 0.3 0.25 Absorbance 0.2 0.15 0.1 0.05 0 0 0.02 0.04 0.06 0.08 Oxidant concentration (mM) Fig 9. Solubility of (nBu)4MnO4 in acetone and acetonitrile as determined by UV spectroscopy. 15 Due to difficulties with finding reproducible conditions for the initial rate experiments, we hypothesized that solubility of the oxidant was rate-limiting in these reactions. Thus, I tested solubility by UV spectroscopy (Figures 8-9) and found that (nBu)4MnO4 was reasonably soluble in both acetone and acetonitrile, while Py2AgMnO4 was only soluble below a threshold of 0.02 mM in acetonitrile and not shown to be soluble above 0.006 mM in acetone. Because the target of this study is ETP precursors which had specifically undergone oxidation using Py2AgMnO4, however, it was elected that the rate experiments would be performed using Py2AgMnO4 in acetonitrile, although the concentration needed ultimately was higher than the limit of solubility at 0.04 mM. To avoid the solubility problem, relative rate measurements were performed. 16 C. RATE DATA Figure 10. Relative rate measurements for the hydantoin library. The rate data collected is presented in Figure 10. From this data, the kinetic isotope effect kH over kD is calculated as 2.26, with a corresponding ∆∆G‡ of 0.534. 17 GC method information: 250 °C inlet, 300 °C detector, flow 1.0 mL/min, oven temperature program: 80 °C for 5 min, 5 °C/min ramp to 150 °C, then 10 °C/min ramp to 300 °C. 1-methylnaphthalene was used as an internal standard. Representative spectra are shown in Figures 11 and 12. Calibrations and rate measurements were done in triplicate. Figure 11. Representative spectrum for calibration. Figure 12. Representative spectrum for obtaining rate data. 18 D. COMPUTATIONAL DATA Values for the parameters which were considered to describe the system are presented in Tables 1-4. Parameters in red were used to create the most descriptive and predictive model (hindrance, the Hammett parameter p of R1, the 13C NMR shift of C5 C5, the NBO charge of N1, and the dihedral angle OH) or as a single parameter to describe the reactivity of the ETP precursors (*R2). Below each table are representations of the parameter meanings (Figures 13-15). Table 1. Calculated parameters to describe IR vibrations and the hydrogen which is abstracted from the hydantoins. C-H C=O  Substrate iC-H iC=O NBOH hindrance 19a 3060.39 8.95 1894.58 93.58 4.57 0.2159 33.64 20a 3038.25 12.71 1881.62 71.64 4.28 0.2131 37.83 21a 3070.16 4.99 1881.72 111.38 5.16 0.2212 38.78 22a 3034.78 5.68 1890.25 125.15 3.95 0.2137 34.56 23a 3029.72 7.16 1877.25 76.39 3.59 0.2114 37.20 24a 3037.20 5.19 1877.47 89.40 4.86 0.2198 43.61 28a 3086.65 10.14 1885.34 59.07 4.54 0.2222 38.50 27a 3067.03 4.71 1884.19 116.16 6.23 0.2326 40.41 29a 3051.38 12.94 1889.08 171.74 4.17 0.2200 36.41 30a 3099.02 17.33 1889.34 141.97 4.67 0.2221 37.38 31a 3052.51 12.90 1874.85 84.55 4.46 0.2181 42.54 32a 3034.56 8.39 1889.75 181.78 4.04 0.2144 36.36 33a 3029.70 10.02 1876.34 79.22 3.67 0.2118 39.57 34a 3072.46 8.00 1878.54 99.24 5.30 0.2227 43.06 35a 3036.99 12.38 1878.32 74.23 4.30 0.2138 38.78 37a 3045.13 12.55 1878.13 88.54 4.43 0.2149 39.31 38a 3033.03 5.61 1884.16 58.31 3.78 0.2142 37.28 39a 3040.13 15.46 1869.86 65.98 4.37 0.2185 37.35 45a 2252.50 7.43 1894.49 93.19 4.28 0.2131 33.05 nC-H hindrance Figure 13. Representative parameters from Table 1. 19 Table 2. Calculated parameters to describe the C5 position of the hydantoins. *R1 p(R1) C5 Substrate LR1 B1R1 B5R1 NBOC5 19a 0.0800 0.1200 3.07 1.70 2.20 70.74 0.13433 20a 0.0800 0.1200 3.08 1.70 2.20 75.63 0.12408 21a 0.0800 0.1200 3.08 1.70 2.20 73.15 0.12297 22a 0.0020 0.1760 4.34 2.07 3.34 81.75 0.12788 23a 0.0020 0.1760 4.36 2.08 3.35 89.23 0.12012 24a 0.0020 0.1760 4.40 2.08 3.35 82.40 0.11791 28a 1.6280 0.2100 8.83 4.31 4.76 84.19 0.13765 27a 2.3040 0.4090 6.37 1.77 3.21 82.45 0.14009 29a 0.0900 0.2040 4.42 2.07 3.37 84.63 -0.1301 30a 0.1110 0.1640 5.39 1.74 4.57 71.8903 0.13372 31a 0.1110 0.1640 5.39 1.73 4.57 75.43 0.12306 32a 1.1540 0.1320 6.41 2.08 3.67 80.13 0.12404 33a 1.0760 0.0870 6.43 2.09 3.69 87.21 0.11777 34a 1.2380 0.1500 6.51 2.16 3.58 83.07 0.11274 35a 0.0800 0.1200 3.08 1.70 2.20 77.21 0.12189 37a 0.0800 0.1200 3.08 1.70 2.20 77.70 0.12409 38a 0.0020 0.1760 4.36 2.07 3.35 88.00 0.11433 39a 0.0020 0.1760 4.35 2.09 3.34 87.37 0.11873 45a 0.0800 0.1200 3.07 1.70 2.20 70.74 0.12408 sp(C5) dC5 Figure 14. Representative parameters from Table 2. 20 Table 3. Calculated parameters to describe the N1 position of the hydantoins and bond angles and distances. *R2 OH Substrate LR2 B1R2 B5R2 NBON1 dCH 19a 0.0000 2.18 1.17 1.17 0.6272 66.79 1.0947 20a 0.0800 3.00 1.70 2.20 0.4378 66.15 1.0969 21a 0.5040 6.45 1.70 3.37 0.4338 64.09 1.0939 22a 0.0000 2.18 1.17 1.17 0.6292 64.61 1.0970 23a 0.0800 3.03 1.70 2.20 0.4434 63.40 1.0979 24a 0.6120 6.46 1.70 3.32 0.4482 57.96 1.0931 28a 0.1340 9.72 3.12 4.23 0.4298 55.58 1.0924 27a 0.7200 6.29 1.77 3.22 0.4255 60.02 1.0942 29a 0.0000 2.17 1.17 1.17 0.6323 62.65 1.0956 30a 0.0000 2.18 1.17 1.17 0.6224 57.68 1.0921 31a 0.0800 3.01 1.70 2.21 0.4351 61.87 1.0955 32a 0.0000 2.17 1.17 1.17 0.6285 64.31 1.0966 33a 0.0800 3.03 1.70 2.21 0.4426 64.12 1.0977 34a 0.6120 6.45 1.70 3.33 0.4483 61.52 1.0939 35a 0.1110 5.26 1.73 4.58 0.4407 67.13 1.0970 37a 0.2940 5.47 1.72 6.16 0.4292 64.20 1.0960 38a 1.4100 4.53 1.70 3.85 0.4427 61.81 1.0974 39a 0.2940 5.20 1.73 6.11 0.4307 62.56 1.0966 45a 0.0000 2.18 1.17 1.17 0.4378 66.79 1.0947 s*N1 NBON1 qOH Figure 15. Representative parameters from Table 3. 21 Table 4. Calculated parameters to describe unoxidized hydantoins and diketopiperazines. Substrate hindrance p(R1) C5 *R2 NBON1 CH 40a 38.64 0.1200 72.81 1.8200 0.4862 63.07 41a 39.98 0.1200 73.88 1.7790 0.4873 58.96 42a 40.26 0.1760 82.96 1.8200 0.5043 55.67 43a 39.06 0.4380 80.84 1.8200 0.4577 59.62 44a 43.55 0.1200 77.05 2.0340 0.4883 60.67 Site 1 1c 44.46 0.0800 73.4965 0.0800 0.4276 108.38 2c 32.63 0.1200 79.534 0.0800 0.4240 34.85 3c 34.27 0.0000 70.7167 0.0800 0.4286 98.01 4c 35.15 0.0000 72.4311 0.0800 0.4292 98.71 5c 43.46 0.0250 77.1527 0.0800 0.4265 108.07 6c 32.32 0.0000 69.7341 0.0800 0.4402 79.21 7c 42.56 0.1200 77.7966 0.0800 0.4194 35.33 8c 45.90 0.1200 73.4965 1.8200 0.4985 115.69 Site 2 1c 33.87 0.2990 79.8218 0.6030 0.4421 96.18 2c 41.63 0.2780 78.6419 0.5190 0.4444 -87.76 3c 34.88 0.4090 89.4857 0.6330 0.4388 94.29 4c 41.34 0.2990 78.5278 1.0350 0.4381 96.91 5c 42.50 0.2990 79.3981 0.6030 0.4359 97.19 6c 40.94 0.0000 77.2317 1.7790 0.4939 -29.67 7c 53.42 0.3880 82.8696 1.0890 0.4410 -75.99 8c 31.37 0.2990 79.9969 1.0890 0.4436 95.39 E. MODELING DATA Using the measured rates and calculated parameters, a model was obtained which correlated the two measurements. The measured and predicted ∆∆G ‡ for this model, as well as the predicted ∆∆G‡ when employing the leave-one-out analysis, are shown in Table 5, and the model along with its coefficients, R 2, and Q2 are shown in Figure 16. From these measurements, the KIE (between substrates 45a, which is deuterated, and 20a, which is not) is computationally calculated as 1.87, and the ∆∆G ‡ is calculated as 0.376. 22 Table 5. Measured and predicted ∆∆G‡ using multivariate correlation and a leave-one-out-analysis. Substrate Measured ∆∆G‡ Predicted ∆∆G‡ Predicted ∆∆G‡ (kcal/mol) (kcal/mol) (LOO) (kcal/mol) 19a 0.0000 0.3245 0.6233 20a 0.9457 1.0405 1.0687 21a 1.2032 0.9225 0.8196 22a 1.0728 1.0733 1.0736 23a 1.9389 2.1811 2.2849 24a 1.2604 1.2060 1.1422 28a 1.2950 1.4249 1.7307 27a 0.3246 0.3180 0.3161 29a 1.0847 1.1473 1.1895 30a 0.0610 0.5856 0.5856 31a 0.6899 0.7495 0.7805 32a 0.4589 0.0412 -0.2086 33a 1.1256 1.0779 1.0571 34a 0.0602 0.3385 0.5055 35a 0.8166 0.9103 0.9434 37a 1.3784 1.1182 1.0771 38a 2.5774 2.2109 2.0839 39a 2.0540 2.1786 2.2203 Validation 45a 1.4793 1.4168 Figure 16. a) five parameter model and b) Leave-one-out plot for five-parameter model. 23 A model was also generated with the exception of both the N 1 NBO charge and the R2=H hydantoin subset. The measured and predicted ∆∆G‡ for this model, as well as the predicted ∆∆G‡ when employing the leave-one-out analysis, are shown in Table 6, and the model along with its coefficients, R2, and Q2 are shown in Figure 17. Table 6. Measured and predicted ∆∆G‡ for the model excluding the N1 NBO charge and hydantoins with a N1-H substitution. Substrate Measured ∆∆G‡ Predicted ∆∆G‡ Predicted ∆∆G‡ (kcal/mol) (kcal/mol) (LOO) (kcal/mol) 20a 0.9457 1.1447 1.2287 21a 1.2032 1.0492 0.9648 23a 1.9389 2.1773 2.2996 24a 1.2604 1.2069 1.1213 28a 1.2950 1.4720 1.9196 27a 0.3246 0.1888 0.0228 31a 0.6899 0.7576 0.7946 33a 1.1256 0.9642 0.8676 34a 0.0602 0.2346 0.3637 35a 0.8166 0.9514 1.0046 37a 1.3784 1.1259 1.0851 38a 2.5774 2.2498 2.1130 39a 2.0540 2.1477 2.1803 Figure 17. a) Model excluding N1-H substitution and N1 NBO charge and b) leave-oneout plot for this model. 24 DISCUSSION Following construction of the models, we sought to relate the relative oxidation rates of the model substrates to the site selectivity of the oxidation of ETP precursors, which contain two distinct C‒H bonds which can undergo oxidation (Fig 7a). To accomplish this, competitive rate measurements were performed between a training set of hydantoins with differing R1 and R2 groups and hydantoin 19a (Figure 10). A relative rate constant krel was obtained for each substrate, and this was converted to a transition state energy difference of ∆∆G‡. The ∆∆G‡ measurements were then correlated to several electronic and steric molecular descriptors using linear regression modeling to quantitatively analyze the substituent effects on the oxidation rate. Figure 18. Model comparing predicted ∆∆G‡ to measured ∆∆G‡. Correlation of computationally derived parameters with the rate measurements led to a rather complex model with five unique terms (Figure 18). These include the Hammett p parameter of the carbon substituent (p(R1)), the NBO charge of the nitrogen 25 adjacent to the hydrogen to be abstracted (NBO N1), the dihedral angle between the hydrogen to be abstracted and the carbonyl C=O bond (OH), the 13C NMR shift of the carbon that is oxidized (C5), and the percent hindrance of the abstracted hydrogen (hindrance). An internal validation employing a leave-one-out analysis (Q2=0.75) suggests a relatively robust model. Considering the model’s complexity, the model’s terms were analyzed individually to gain understanding about which factors impact the oxidation’s selectivity. It was hoped that by doing so, the selectivity of the oxidant on the ETP precursors would be better understood. By analyzing single-parameter correlations within subsets with either consistent R1 or R2 substituents, notable trends were discovered among subsets of the hydantoins. Two of the most significant parameters as determined by their coefficients are p(R1) and hindrance. In considering p(R1), it is clear that hydantoins with the same R 1 substituent cluster together. However, R2=phenyl substituents are an outlier with a much higher p(R1) in each set (Figure 19a,b), corresponding with a moderately low ∆∆G‡ for these substrates. Because p describes resonance and electronic effects, it is reasonable that phenyl would have a greater impact on the models than the remainder of the substituents, with which nitrogen cannot resonate. Using this knowledge, hindrance was examined by excluding R2=phenyl substrates from subsets in which R1 did not change. By employing this method, it was found that hindrance due to R 2 increases the energy barrier of the reaction (Figure 19c,d). This makes intuitive sense because sterically limiting the ability of the oxidant to reach the hydrogen which is abstracted should reduce the reaction rate for a mechanism in which hydrogen atom abstraction is rate limiting. 26 Figure 19. Descriptions of variations in the electronic (a,b) and steric (c,d) contributions of the N1 substituent to the model. While p(R1) and hindrance elucidate the effects of the R2 substituent, variation in R1 is described by examining the 13C NMR shift of C5. Holding the R2 substituent constant, several trends were apparent. For the R2=H and R2=Me subsets, the energy barrier increases as a function of increasing NMR chemical shift of C5 (Figure 20). Greater electron density thus appears to stabilize radical formation. This is in agreement with external KIE studies performed by my collaborator that suggest hydrogen atom abstraction is rate limiting. This trend is perturbed, however, by hydantoins with R2=phenyl. In the case of these substrates, no clear trend is present. This suggests that the 27 contributions of the phenyl ring to the electronic nature of C5 override the effects of the carbon substituent. Figure 20. The 13C NMR shift of C5 describes variation in the C5 substituent. The N1 NBO charge and dihedral angle play a necessary compensatory role to the more impactful parameters. When R2 substituents were compared, the R2=H substituted hydantoins were found to have a much lower N 1 NBO charge than the remainder of the hydantoins, accompanied by high reactivity (Figure 21). Generation of a model excluding the R2=H subset resulted in a comparable model to that in Figure 1, with an R2 of 0.92 and a Q2 of 0.79 (Fig. 17) with the exclusion of the N1 NBO charge. This parameter thus functions solely to normalize the R2=H subset. Using a similar analysis, the importance 28 of the dihedral angle was probed, but the only trend that could be found was a general clustering together of hydantoins with the same R1 substituent. Thus, it may compensate for effects not described by the other four descriptors. Figure 21. Contributions of the N1 NBO charge and the dihedral angle to the complex model. In summary, p(R1), the N1 NBO charge, and hindrance describe variation in R2, while C5 describes variation in R1. The 13C NMR shift of C5 shows that deshielding C5, which is connected the hydrogen which is abstracted, decreases the rate of reaction. Hindrance plays an intuitive role of showing that the more sterically hindered the oxidation site is, the higher the barrier to reaction is, while p and the NBO charge function to describe the R2 substituents Ph and H, respectively. While the analysis of these subsets provides information about the important factors determining selectivity of this oxidation, the comprehensive model can be validated by determining its ability to predict. As an external validation, the model’s ability to predict an internal kinetic isotope effect was determined. The experimentally determined difference in H versus D is 0.38 29 kcal/mol, which is in reasonable agreement with the computational model prediction of 0.53 kcal/mol (Figure 18). This corresponds to a calculated KIE of 1.9 (2.3 experimentally) and is the first example of such a prediction using these multivariate analysis tools. We then turned to the initial goal of the study, using the mechanistic information gained to determine how the oxidant achieved selectivity among the ETP precursors. To do this, we envisioned using the model value of ∆∆G ‡ to predict simply whether or not an oxidation would occur for an untested substrate. In order to attempt this, the parameters used in the model were collected for the ETP precursors previously submitted to the oxidation conditions in the Movassaghi group’s natural product syntheses. Each of these eight diketopiperazines had two distinct potential sites of oxidation, and of these, 11 sites were successfully oxidized and five were not oxidized (Figure 7b). Using the mathematical model above, combined with specific descriptors for the diketopiperazines, ten of the eleven sites that experimentally undergo oxidation have low predicted energy barriers of <1 (Figure 22a). Of the five that do not readily oxidize, three have relatively large energy barriers of >6 while another is predicted moderately high at 2.65. In this significant extrapolation of the simple model system, 13 out the 16 sites are successfully predicted with one moderately predicted and two errors. Examining the hydantoins which were and were not well predicted, the model seems well-equipped to predict proximal steric effects, but falls short of predicting electronic or more distant effects. 30 Figure 22. Examples of full model and single parameter predictions of which diketopiperazine sites would undergo oxidation. Although the complex model accurately predicted most of the diketopiperazine oxidations, it has several drawbacks. As well as being complex, it failed to accurately predict that several hydantoins, 40a-44a, would not undergo oxidation. Thus, the parameters were examined individually to determine whether any single parameter yielded a threshold over which hydantoins and diketopiperazines were not oxidized. Employing this method, it was discovered that *R2 had just such a threshold; hydantoins and diketopiperazines with a *R2 above 1.04 were not oxidized, while those below were oxidized (Figure 22b). This pattern held for all oxidized and unoxidized hydantoins, as 31 well as 13 of the 16 diketopiperazine reactive sites. The Taft parameter * is predominantly a measure of the polar nature of a substituent, where higher polarizability corresponds to a higher value. This suggests that variation in the electronic nature of the substrate as dictated by the R2 substituent is the primary determiner of the ability of this oxidation to proceed. This single parameter analysis has the advantage of being faster to per-form and is more accurate than the predictions of the complex model. While *R2 was used to validate whether or not already tested ETP precursors would undergo oxidation, this model could be used to predict whether other precursors of interest would be likely to undergo oxidation. In summary, a class of simple molecules has been used to examine the Py2AgMnO4 oxidation of tertiary hydrogens to form quaternary centers, and it has been shown that C–H bond cleavage is the rate determining step. Characterizing which substrate features influence this step resulted in the identification of a model which accurately predicted whether or not most ETP precursors would undergo oxidation using the permanganate oxidant. These findings may be used to predict which precursors to ETPs are promising targets for undergoing the oxidative step of future syntheses. 32 REFERENCES (1) Jaffé, H, H. “A Reexamination of the Hammett Equation.” Chem. Rev. 1953, 53, 191. (2) (a) Taft, R. W. “Polar and Steric Substituent Constants for Aliphatic and o-Benzoate Groups from Rates of Esterification and Hydrolysis of Esters” J. Am. Chem. Soc. 1952, 74, 3120. b) Taft, R. W. “Linear Free Energy Relationships from Rates of Esterification and Hydrolysis of Aliphatic and Ortho-substituted Benzoate Esters” J. Am. Chem. Soc. 1952, 74, 2729. (3) (a) Toste, F. D., Sigman, M. S., Miller, S. J. "Pursuit of Noncovalent Interactions for Strategic Site-Selective Catalysis" Acc. Chem. Res. 2017, 50, 609. b) Sigman, M. S.; Harper, K. C.; Bess, E. N.; Milo, A. "The Development of Multidimensional Analysis Tools for Asymmetric Catalysis and Beyond." Acc. Chem. Res., 2016, 49, 1292–1301. (4) (a) Neel, A. J.; Milo, A.; Sigman, M. S.; Toste, F. D. Enantiodivergent Fluorination of Allylic Alcohols: Dataset Design Reveals Structural Interplay Between Achiral Directing Group and Chiral Anion. J. Am. Chem. Soc. 2016, 138, 3863. (b) Milo, A.; Neel, A. J.; Toste, F. D.; Sigman, M. S. “A Data-Driven Approach to Mechanistic Elucidation in Chiral Anion Catalysis.” Science 2015, 347, 737. (c) Santiago, C. B.; Milo, A; Sigman, M. S. "Developing a Modern Approach To Account for Steric Effects in Hammett-Type Correlations," J. Am. Chem. Soc. 2016, 138, 13424. (5) Neel, A. J.; Hilton, M. J.; Sigman, M. S.; Toste, F. D. "Exploiting non-covalent π interactions for catalyst design" Nature 2017, 543, 637. (6) Niemeyer, Z.; Milo, A.; Hickey, D. P.; Sigman, M. S. "Parameterization of phosphine ligands reveals mechanistic pathways and predicts reaction outcomes," Nature Chem. 2016, 8, 610. 33 (7) (a) Neel, A. J.; Hehn, J. P.; Tripet, P. E.; Toste, F. D. “Asymmetric CrossDehydrogenative Coupling Enabled by the Design and Application of Chiral-Triazole Containing Phosphoric Acids.” J. Am. Chem. Soc. 2013, 135, 14044. (b) Phipps, R. J.; Hamilton, G. L.; Toste, F. D. “The Progession of Chiral Anions From Concepts to Applications in Asymmetric Catalysis.” Nat. Chem. 2012, 4, 603. (8) Woods, B. P.; Orlandi, M.; Huang, C.-Y.; Sigman, M. S.; Doyle, A. G. “NickelCatalyzed Enantioselective Reductive Cross-Coupling of Styrenyl Aziridines” J. Am. Chem. Soc. 2017 ASAP (9) Sevov, C. S.; Hickey, D. P.; Cook, M. E.; Robinson, S. G.; Barnett, S.; Minteer, S. D.; Sigman, M. S.; Sanford, M. S. " Physical Organic Approach to Persistent, Cyclable, LowPotential Electrolytes for Flow Battery Applications," J. Am. Chem. Soc. 2017, 139, 2924. (10) (a) Weldy, N. M.; Schafer, A. G.; Owens, C. P.; Herting, C. J.; Varela-Alvarez, A.; Chen, S.; Niemeyer, Z.; Musaev, D. G.; Sigman, M. S.; Davies, H. M. L.; Blakey, S. B. "Iridium(III)-bis(imidazolinyl)phenyl catalysts for enantioselective C–H functionalization with ethyl diazoacetate," Chem. Sci. 2016, 7, 3142–3146. (b) Mougel, V.; Santiago, C. B.; Zhizhko, P. A.; Bess, E. N.; Varga, J.; Frater, G.; Sigman, M. S.; Copéret, C. “Quantitatively Analyzing Metathesis Catalyst Activity and Structural Features in SilicaSupported Tungsten Imido–Alkylidene Complexes,” J. Am. Chem. Soc. 2015, 137, 66996704. (11) Zheng, C.-J.; Kim C.-J.; Bae, K.S.; Kim Y.-H.; Kim W.-G. “Bionectins A-C, epidithiodioxopiperazines with anti-MRSA activity, from Bionectra byssicola F120” J. Nat. Prod. 2006, 69, 1816. 34 (12) Jiang, C.-S.; Guo, Y.-W. “Epipolythiodioxopiperazines from fungi: chemistry and bioactivities.” Mini Rev. Med. Chem. 2011, 11, 728. (13) (a) Rightsel, W. A.; Schneider, H. G.; Sloan, B. J.; Graf, P. R.; Miller, F. A.; Bartz, Q. R.; Ehrlich, J.; Dixon, G. J. “Antiviral activity of gliotoxin and gliotoxin acetate.” Nature. 1968, 204, 1333. (b) Miller, P. A.; Milstrey, K. P.; Trown, P. W. “Specific inhibition of viral ribonucleic acid replication by Gliotoxin.” Science. 1968, 159, 431 (14) Coleman, J. J.; Ghosh, S.; Okoli, I.; Mylonakis, E. “Antifungal activity of microbial secondary metabolites” PLoS ONE. 2011, 6, e25321. (b) Speth, C.; Kupfahl, C.; Pfaller, K.; Hagleitner, M.; Deutinger, M.; Würzner, R.; Mohsenipour, I.; Lass-Flörl, C.; Rambach, G.; “Gliotoxin as putative virulence factor and immunotherapeutic target in a cell culture model of cerebral aspergillosis”. Mol. Immunol. 2011, 48, 2122. (15) (a) Waring, P.; Beaver, J. “Gliotoxin and related epipolythiodioxopiperazines. Gen. Pharmac.” 2012, 27, 1311. (b) Müllbacher, A.; Waring, P.; Tiwari-Palni, U.; Eichner, R. D. “Structural relationship of epipolythiodioxopiperazines and their immunomodulating activity.” Molec. Immunol. 2005, 23, 231. (16) (a) Kim, J.; Movassaghi, M. “Biogenetically-Inspired Total Synthesis of Epidithiodiketopiperazines and Related Alkaloids.” Acc. Chem. Res. 2015, 48, 1159. (b) Boyer, N.; Morrison, K. C.; Kim, J.; Hergenrother, P. J.; Movassaghi, M. “Synthesis and Anticancer Activity of Epipolythiodiketopiperazine Alkaloids ” Chem. Sci. 2013, 4, 1646. (17) (a) Coste, A.; Kim, J.; Adams, T. C.; Movassaghi, M. “Concise Total Synthesis of (+)-Bionectins A and C” Chem. Sci. 2013, 4, 3191. (b) Boyer, N.; Movassaghi, 35 M. “Concise Total Synthesis of (+)-Gliocladins B and C.” Chem. Sci. 2012, 3, 1798. (c) Kim, J.; Movassaghi, M. “General Approach to Epipolythiodiketopiperazine Alkaloids: Total Synthesis of (+)-Chaetocins A and C and (+)-12,12'-Dideoxychetracin A.” J. Am. Chem. Soc. 2010, 132, 14376. (18) Gaussian 09, Revision A.02, M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A. Marenich, J. Bloino, B. G. Janesko, R. Gomperts, B. Mennucci, H. P. Hratchian, J. V. Ortiz, A. F. Izmaylov, J. L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini, F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V. G. Zakrzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, K. Throssell, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro, M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, T. Keith, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, J. M. Millam, M. Klene, C. Adamo, R. Cammi, J. W. Ochterski, R. L. Martin, K. Morokuma, O. Farkas, J. B. Foresman, and D. J. Fox, Gaussian, Inc., Wallingford (19) Zhao, Y.; Truhlar, D. G.; “The M06 Suite of Density Functionals for Main Group Thermochemistry, Thermochemical Kinetics, Noncovalent Interactions, Excited States, and Transition Elements: Two New Functionals and Systematic Testing of Four M06 Functionals and Twelve Other Functionals.” Theoretical Chemistry Accounts 2008, 120, 215 36 (20) Weigend, F.; and Ahlrichs, R. “Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy.” Phys. Chem. Chem. Phys. 2005, 7, 3297. (21) Quinn, J. A. Molecular Modeling Pro, ChemSW, Inc.: Fairfield , CA. (22) MATLAB Student Version, The MathWorks, Inc.: Natick, MA, 2015. (23) Seeman, J. L. “The Curtin-Hammett principle and the Winstein-Holness equation: new definition and recent extensions to classical concepts.” J. Chem. Educ. 1986, 63, 42.