Enhancing chemoselectivity and predicting site-selectivity in manganese catalyzed C–H oxidation

Ahn, Chiyoung

Permalink

https://hdl.handle.net/2142/129656 Copy

Description

Title

Enhancing chemoselectivity and predicting site-selectivity in manganese catalyzed C–H oxidation

Author(s)

Ahn, Chiyoung

Issue Date

2024-12-20

Director of Research (if dissertation) or Advisor (if thesis)

White, M. Christina

Doctoral Committee Chair(s)

White, M. Christina

Committee Member(s)

• Hergenrother, Paul J.
• van der Donk, Wilfred A.
• Snyder, Benjamin E.R.

Department of Study

Chemistry

Discipline

Chemistry

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• chemoselective remote C-H oxidation
• metal-oxo
• PDP catalysis
• neural network
• site-selectivity prediction

Abstract

Recent advancements in Csp³–H oxidation methods have shown that even strong, oxidatively inert C–H bonds can be selectively oxidized in the presence of more oxidatively labile functional groups. Remarkably, these methods enable preferential oxidation of specific C–H bonds within complex substrates, despite minimal inherent reactivity differences among them. These breakthroughs have significantly expanded the utility of Csp³–H oxidation in total synthesis and late-stage metabolite synthesis. Notably, PDP-catalyzed remote C–H oxidation has demonstrated that specific sites of oxidation can be preferentially targeted based on electronic, steric, and stereoelectronic selection rules. Additionally, the chemoselective oxidation of C–H bonds even in the presence of traditionally more oxidatively labile functional groups, such as electron-neutral aromatics, has been achieved. As the application of PDP-catalyzed C–H oxidation extends to more complex substrates, two main challenges have emerged: improving chemoselectivity and accurately predicting site of oxidation. The highly reactive carboxylate-oxo active oxidant in PDP catalysis enables strong C–H bond oxidation but remains incompatible with many functional groups. Moreover, increased molecular complexity introduces a greater number of potential oxidation sites, making heuristic site-selectivity rules harder to interpret due to multiple conflicting factors compared to simpler substrates previously used to define these rules. This uncertainty in reaction outcomes significantly limits the broader synthetic application of this chemistry. This dissertation presents efforts toward enhancing chemoselectivity and predicting site selectivity in PDP-catalyzed C–H oxidation chemistry. Firstly, improvements in chemoselectivity were achieved, resulting in the tolerance of α,β-unsaturated carbonyls prevalent in bioactive natural products. By removing the carboxylic acid additive and changing the solvent from acetonitrile to 1,1,1,3,3,3-hexafluoroisopropanol (HFIP), a more charged pathway was established, improving chemoselectivity for C–H oxidation over epoxidation of α,β-unsaturated carbonyls. This was confirmed by competitive kinetic studies. While inductive deactivation through solvent hydrogen bonding was beneficial, it was insufficient alone to achieve high chemoselectivity. Additionally, a novel catalyst, Mn(MeCF₃-PDP), was developed, demonstrating superior performance in oxidizing complex natural products compared to the previous catalyst Mn(CF₃-PDP), in terms of yield and mass balance due to steric restrictions in the catalyst scaffold. Secondly, a neural network–based quantitative site prediction model was developed. A novel steric descriptor was introduced, capturing catalyst–C–H interactions to provide a holistic view of the steric landscape of the target C–H bond. This contrasts with previous steric measurements based on connectivity, substituent size, or oversimplified probes. Furthermore, to enable the model to discern subtle differences between potentially oxidizable methylene and methine sites, novel neural network–based data preprocessing modules were designed based on chemical principle guided constraints. This architectural design allowed the model to rapidly learn subtle differences between chemically similar remote C–H sites, achieving an average prediction accuracy of 94% in K-fold validation. These advancements contribute to the broader application of PDP-catalyzed C–H oxidation in complex synthetic settings by enhancing chemoselectivity and providing a reliable tool for predicting site selectivity.

Graduation Semester

2025-05

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/129656

Copyright and License Information

Copyright 2025 Chiyoung Ahn

Owning Collections