Researchers at Osaka University have used machine learning to develop a boron-based catalyst that is highly effective at chemically converting amino acids and peptides, producing only water as a byproduct.
Imagine synthesizing and testing over 50 different complex molecules to identify the most effective catalyst for a particular chemical reaction. Traditional approaches to developing new catalysts for chemical reactions in this “try and see” manner are often very labor-intensive and require repeating numerous experiments with potential candidate molecules. . Machine learning techniques, which are now widely used, can be used to perform this task more efficiently by predicting catalyst performance in advance based on theoretical properties.
In a groundbreaking study published in Nature Communications, researchers at Osaka University used a computer library of synthesized molecules, along with molecules that are currently entirely theoretical, to find the best catalyst for a specific chemical reaction. I used
The goal of the research was to find a better way to add carbon groups to amino acids and peptides, which are very common in living things, to change the properties of these compounds. Like many reactions, these processes are facilitated by catalysts, but traditional metal-based catalysts are often toxic and expensive. Although this work aimed to use triarylborane as a catalyst, there are potentially hundreds of possibilities due to its relatively complex structure. These compounds are based on boron, a major element that is relatively cheap and has low toxicity.
“Evaluating molecular catalysts for organic synthesis can be very time-consuming,” says Yuki Hisada, lead author of the study. “In the case of the triarylborane used in our study, there are many permutations of molecular structure, and just identifying the best candidate can require months of research.”
The researchers combined experimental data from a limited number of synthesized triarylboranes with properties predicted for other molecules not yet synthesized using theoretical calculations to create a list of 54 possible catalysts. I created a library.
“In this process, we evaluated parameters predicted to influence the progress of the reaction,” explains corresponding author Yoichi Hoshimoto. “These include factors such as molecular orbital energy levels and energy barriers to specific processes.”
Promising candidates were identified using Gaussian process regression using an in silico library, and testing with this triarylborane demonstrated a high level of performance. This compound facilitates the reaction of amino acids with very high yields and tolerates the presence of a large number of different functional groups. As an additional advantage, these reactions successfully use molecular hydrogen H2 as a reagent, thus producing only water as a harmless by-product.
The study also looked at other ways to reduce the environmental impact of the process and found that the dangerous solvent tetrahydrofuran could be replaced with a less toxic alternative, 4-methyltetrahydropyran.
Modern chemists face increasing demands, juggling the development of new syntheses with limited allies while considering environmental impact, efficiency, cost, sustainability, and other factors. This research demonstrates important advances in the use of machine learning to streamline the development of new chemical processes and how these new processes incorporate changes that work together to produce green systems. We emphasize what you can do.
Figure 1 Challenges to direct functionalization of amino acids/peptides
Provided by: Osaka University
Figure 2 This study: boron-catalyzed direct alkylation of amino acids/peptides, producing water as the only waste product.
Provided by: Osaka University
Figure 3 Harmonizing machine learning and experimentation to accelerate the catalyst optimization process
Credit: Osaka University
The paper “In silico-assisted derivatization of triarylboranes for catalytic reductive functionalization of aniline-derived amino acids and peptides with H2” was published in Nature Communications at DOI: https://doi.org/10.1038/S41467- 024 -47984-0.
/Open to the public. This material from the original organization/author may be of a contemporary nature and has been edited for clarity, style, and length. Mirage.News does not take any institutional position or position, and all views, positions, and conclusions expressed herein are those of the authors alone. Read the full text here.
