Supramolecular Catalyzed Cascade Reduction of Azaarenes Interrogated via Data Science

Abstract

Catalysis of multicomponent transformations requires controlled assembly of reactants within the active site. Supramolecular scaffolds possess synthetic microenvironments that enable precise modulation over noncovalent interactions (NCIs) engaged by reactive, encapsulated species. While molecular properties that describe the behavior of single guests in host cavities have been studied extensively, multicomponent transformations remain challenging to design and deploy. Here, simple univariate regression and threshold analyses are employed to model reactivity in a cascade reduction of azaarenes catalyzed by water-soluble metal organic cages. Yield and stereoselectivity models help deduce unknown mechanisms of reactivity by the multicomponent, host-guest complexes. Furthermore, a comprehensive model is established for NCIs driving stereoselectivity in the reported host-guest adducts.

Figure 1

Investigating reactive, ternary complexes requires chemically coherent systems and models. (A) Metal catecholate cages possess synthetically tractable, biomimetic binding pockets. (B) Data science helps generate chemical knowledge based on correlations between chemically meaningful descriptors and experimental observables. (C) Proposed model azaarene reduction cascade to interrogate ternary complex formation and reactivity.

Figure 2

Establishing the host-mediated azaarene reduction cascade reactivity, scope, and mechanism. (A) Reaction discovery and controls. (B) Systematic series of quinolines, see Table S6, ^a4:1 dr, ^b19:1 dr, and ^c9:1 dr. (C) Representation of molecular volume in its most compact conformer linearly correlates (R² = 0.77) with yields of supramolecular catalyzed transformation and are well represented by a univariate threshold analysis (yield threshold = 30%, detected molecular volume threshold = 171.7 ų, MCC = 0.83).

Figure 3

Chemical principles describing reactivity of ternary complexes transfers across distinct hosts assemblies. (A) Exchange of the metal cation of the supramolecular cage increases the flexibility of the cavity, leading to differences in guest exchange rates. Using a pyrene ligand architecture increases the size of the cavity and enables more facile guest exchange. (B) Minimum molecular volume maintains robust linear correlations across the catalyst series (up to R² = 0.84), with corresponding excellent univariate threshold analyses (yield threshold = 30%, up to MCC = 1.0), see Tables S6–S9.

Figure 4

Reactivity of the ternary complex over an expanded substrate scope divergently correlates to substrate stereoelectronic features. (A) Additional quinolines scope and limited prediction capacity of minimum molecular volume on reaction yield, see Table S10; ^a1.7:1, dr, ^b11.2:1 dr, ^c10.5:1 dr, ^d4.6:1 dr, and ^e4.5:1 dr. (B) Unsupervised clustering reveals two classes of quinoline substrates. (C) Cluster specific bivariate linear regression using molecular volume and molecular dipole differentially correlates to reaction yield, up to R² = 0.80, see Table S11.

Figure 5

Evaluation of transient, stereoselective interactions across the reactive, ternary complex within the host-mediated azaarene reduction cascade. (A) Sterimol BMin and electronegativity of the quinoline substrate correlate to the diastereoselectivity of the azaarene reduction, R² = 0.77, see Table S14. (B) Buried volume at C-4 of pyridine–borane correlates to differences in diastereoselectivity of the 1j reduction, R² = 0.90, see Table S15. (C) NCIs across proposed selectivity determining ternary complex recapitulates observed diastereoselectivity for a series of quinolines with 2a, R² = 0.87, see Table S17. (D) NCIs across proposed selectivity determining ternary complex recapitulates observed diastereoselectivity for a series of pyridine–boranes with 1j, R² = 0.96, see Table S18. (E) Comprehensive model for the NCIs driving stereoselectivity in the reported reactive, ternary host–guest adducts.