NMR Landscapes for Chemical Shift Prediction

Abstract

The ability to reliably predict NMR chemical shifts plays an important role in elucidating the structure of organic molecules. Additionally, an intriguing question is how the multitude of variable factors (structural, electronic, and environmental) correlate with the actual electromagnetic shielding effect that determines the chemical shift value. This work presents NMRscape as a new tool for understanding these correlations by constructing the landscape that describes the relationship between the chemical shift value and the moieties bonded to a molecular scaffold. The scaffold may be as small as a single atom probed by NMR or a larger molecular framework containing the probed atom. NMRscape operates with only a list of the chemical moieties bonded to the scaffold, without utilizing any potentially biasing chemometric descriptors. The corresponding chemical shift landscape is constructed based on fundamental physical principles, which makes NMRscape a credible chemical shift prediction and analysis tool. As an illustration, we demonstrate that NMRscape can predict (13)C chemical shifts with an accuracy exceeding the substituent chemical shift (SCS) increment, hierarchical organization of spherical environments (HOSE), and neural networks (NN), methods for three distinct families of molecules sharing a common scaffold structure with moieties placed at two variable sites. The constructed NMR landscapes confirmed known empirical rules relating chemical shift values to the variation of chemical moieties on a scaffold, as well as uncovered hitherto hidden relationships. The practical importance of NMRscape is discussed.

Fundamental Principles of Control Landscapes with Applications to Quantum Mechanics, Chemistry and Evolution

“…For the two-dimensional case with moieties X (rows) and Y (columns), reordering involves permuting the rows and columns in order to identify an ordering O that reveals the most regular property landscape, as exemplified by going from Figure 2.5(a) to 2.5(b) [33,34,36,55]. In these and other studies, the recovered landscapes were trapfree.…”

“…For a family of molecules in a "library" built around a fixed molecular scaffold, the variables can be defined in terms of prescribed integer labels of the chemical moieties on each site. This procedure is illustrated in Figure 2.5 with an example from [36]. The molecular scaffold is a vinyl group, the variable moieties X and Y may take on the integer values 1 through 15 as labeled in Figure 2.5(a), and the objective "property" is spectral, consisting of the 13 C nuclear magnetic resonance (NMR) chemical shift of the indicated carbon atom.…”

“…It has been shown that surfaces that inherently contain traps are generally not transformed into trap-free surfaces under reordering [11], indicating that reordering cannot artificially transform inherently rugged landscapes into smooth ones. The reordering technique has been applied to a number of molecular property objectives [12,28,33,34,36,55].…”

“…Recent work showing that an iterative method can accelerate the identification of optimal molecules or materials [34] and facilitate accurate NMR spectral peak prediction [36] represents just the beginning of many potential applications of OptiChem theory. Additionally, the understanding that a well-posed chemical optimization objective should produce a trap-free control landscape could be important for designing effective optimization search algorithms, as well as taking advantage of well-known empirical "rules" of chemistry [38].…”

“…In such cases, OptiChem theory may be used to prevent the waste of experimental resources in a suboptimal portion of the available search space. Second, individual outliers on the landscape may indicate errors in the experimental data, as was shown in [36], where the outlier nature of the associated data points was only revealed upon applying the landscape principles. Outlier points on the landscape may also reveal previously hidden distinct physical mechanisms underlying the control objective, as was found in a study optimizing photoluminescent quantum yield in transition metal complexes [33].…”