Tech

Artificial intelligence helps find new natural substances

Martin Hoffmann of the University of Jena has announced COSMIC (Confidence Of Small Molecule IdentifiCations). This is a workflow that makes it faster and easier to identify small molecules than before.Credits: Jens Meyer / University of Jena

More than one-third of all medicines currently available are based on natural active ingredients, and a research team at the University of Jena has developed a procedure to identify small active ingredient molecules much faster and easier. Secondary natural substances found in many plants, bacteria, and fungi have anti-inflammatory properties that can ward off pathogens and even prevent the growth of cancer cells. However, leveraging the wealth provided by natural medicine racks to identify new natural substances is time-consuming, costly and labor-intensive. A team of bioinformaticians at the University of Friedrichschiller Jena has developed a method that allows for much faster and easier identification of small active substance molecules. Researchers have developed their method called COSMIC (Confidence Of Small Molecule IdentifiCations). Nature biotechnology..


Millions of structural data items that have not yet been decrypted

To find out what substances are in a biological sample, such as a plant extract, researchers analyze the sample using: Mass spectrometry.. In this process, the molecules are broken down into fragments and their masses are determined. “Using the CSI: FingerID molecular search engine we developed, Molecular structure It matches these fragments, “says Professor Sebastian Becker of the University of Jena. “Whether this search was successful, that is, whether the search results are correct. structure— It is not something that can be distinguished in this way. “

Currently, there is a vast collection of data containing billions of mass spectrometric data items from the analysis of millions of biological samples, most of which have not been identified for their structure. This is where COSMIC comes in, allowing it to automatically decipher the structure of most unidentified molecules.

“For this purpose, we use machine learning techniques,” explains Martin Hofmann, lead author of the new publication. “First, the mass spectrum of the sample under investigation is compared to the available structural data.” As with Google search, we get a more or less extensive list of potential hits. “Our method shows how confident that the first hit we find is actually the structure we’re looking for,” Hoffman adds. To do this, COSMIC evaluates the quality of the proposed hits and determines a score that infers whether it is right or wrong.

Discovered new bile acids

Becker and his team were able to work with colleagues at the University of California, San Diego to demonstrate how well their method actually works. They studied mass spectrometric data from the digestive system of mice and searched for yet unknown bile acids. To this end, more than 28,000 theoretically possible bile acid structures have been constructed and compared with measured data from the mouse microbiota. Subsequent analysis by COSMIC resulted in a total of 11 new bile acid structures that were previously completely unknown. Two of these have since been confirmed using specially synthesized reference samples.

“This first shows that our method works,” emphasizes Sebastian Becker. Second, COSMIC allows screening to be performed completely automatically in a very short time without any manual labor, greatly accelerating the search for new and interesting substances. Becker anticipates that in the next few years it will be possible to uncover thousands of new molecular structures in this way.


Identifying complex classes by machine learning


For more information:
Martin A. Hoffmann et al, Reliable structural annotation of metabolites not present in the spectral library, Nature biotechnology (2021). DOI: 10.1038 / s41587-021-01045-9

Quote: Artificial intelligence is a new natural substance acquired from https://phys.org/news/2021-10-artificial-intelligence-natural-substances.html on October 14, 2021 (October 14, 2021). Helps you find

This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. The content is provided for informational purposes only.



Artificial intelligence helps find new natural substances

Source link Artificial intelligence helps find new natural substances

Show More

Related Articles

Back to top button