A framework for automated structure elucidation from routine NMR spectra and applications and mechanics of organic reactions in alkali salt media
Abstract/Contents
- Abstract
- Methods to automate structure elucidation of organic molecules have the potential to accelerate chemical discovery and are critical for the broad application of automated high-throughput experimentation platforms. Nuclear magnetic resonance (NMR) spectroscopy is one of the most powerful methods for elucidating unknown structures because NMR spectra have a high information content that encodes the local environments of atoms that make up organic molecules. However, the analysis of NMR data requires a trained chemist and even for experts it is often time-consuming and error-prone. The first part of this dissertation describes the development and initial application of a machine-learning-based method to automate small-molecule structure elucidation from 1D NMR spectra. The use of computers to aid NMR analysis has a long history. Chapter 1 provides an overview of this area, including NMR spectra prediction (the opposite of structure elucidation), computer-assisted structure elucidation (CASE), and efforts to fully automate NMR data analysis and structure elucidation. With this backdrop, Chapter 2 describes a new framework for automated structure elucidation. First, a machine learning model was created that takes experimental 1D 1H and 13C NMR data and a molecular formula as inputs and generates a probabilistic prediction for the presence of almost a thousand substructures. The model was trained on > 100,000 simulated NMR spectra and then validated and tested using a few hundred experimental spectra for relatively small organic molecules. The model is shown to be remarkably accurate for predicting the presence of substructures. An F1 score of 0.803 was obtained for substructure prediction of the experimental test set. It can also annotate the spectra to label the peaks corresponding to specific predicted substructures. Second, a graph generator was developed that takes the substructure probability profile output by the machine learning model and generates a probabilistic ranking of candidate molecular structures (constitutional isomers). Remarkably, for the test dataset of experimental spectra of molecules the algorithm has never seen before, the correct structure was identified 67.4% of the time, and the correct structure was one of the top-ten candidates in 95.8% of the cases. The chapter concludes with an outlook for how to improve and extend the framework to enable structure elucidation of more complex and larger molecules. The second part of the thesis explores organic reactions under solvent-free conditions in media comprised of alkali salts. Compared to solution-phase reactions, there has been substantially less development of solvent-free reactions and even less so for reactions in alkali salts. These unusual media can enable reaction pathways that are difficult to access in conventional solution-phase systems, providing opportunities to discover new methods that streamline synthesis. Chapter 3 describes the investigation of hydrocarboxylation reactions via the addition of alkali formate to alkenes in alkali salt media at elevated temperatures. Solvent-free conditions for high-yielding hydrocarboxylations were developed for various , -unsaturated alkene carboxylates and a few simple alkenes. Mechanistic studies support a radical-mediated pathway for these reactions. With this insight, alkali formate addition to alkenes was translated to aqueous solution-phase conditions using water-soluble radical initiators. Chapter 4 describes the addition of hypophosphite to alkenes in solvent-free alkali salts. The addition of hypophosphite shares many similarities with the addition of formate but has a considerably broader substrate scope, including both , -unsaturated carboxylates and unconjugated alkene carboxylates. Mechanistic studies and product distributions are consistent with a radical mechanism, and many of the reactions can be performed under aqueous conditions with radical initiators. Several products that are difficult to obtain with other methods were synthesized, simplifying the synthesis of polyfunctional H-phosphinates. Finally, Chapter 5 discusses several additional solvent-free reactions in alkali salts at an earlier stage of development. Reactions include the addition of acetate to alkenes, the conversation of glycolate to tricarballylate, a multistep cycloaddition between maleate and glycine, and aromatic dehalogenation.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2022; ©2022 |
| Publication date | 2022; 2022 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Huang, Zhaorui |
|---|---|
| Degree supervisor | Kanan, Matthew William, 1978- |
| Thesis advisor | Kanan, Matthew William, 1978- |
| Thesis advisor | Burns, Noah |
| Thesis advisor | Du Bois, Justin |
| Degree committee member | Burns, Noah |
| Degree committee member | Du Bois, Justin |
| Associated with | Stanford University, Department of Chemistry |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Zhaorui Huang. |
|---|---|
| Note | Submitted to the Department of Chemistry. |
| Thesis | Thesis Ph.D. Stanford University 2022. |
| Location | https://purl.stanford.edu/pp142wz2062 |
Access conditions
- Copyright
- © 2022 by Zhaorui Huang
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...