Only through machine learning and AI can we understand all the chemicals that surround us, say the researchers.

Machine Learning


Only through machine learning and AI can we understand all the chemicals around us

Illustration of the problem. Of the vast amount of molecules in chemical space, current technology can only detect a limited amount. An even smaller percentage of molecules are actually identified. The exposome chemical space, i.e., the molecules we are exposed to, far exceeds these realms of measurable, measured, and identified molecules. Credit: HIMS/JACS.

of Jack's O presented an invited perspective by Dr Saer Samanipour and his team on the daunting task of mapping all the chemicals around us. Samanipour, an assistant professor at the Van 't Hoff Institute of Molecular Science at the University of Amsterdam (UvA), takes stock of the available science and concludes that truly proactive chemical management is currently not feasible.

To truly get a grip on the vast and ever-expanding world of chemistry, Samanipour advocates for the use of machine learning and AI to complement existing strategies for detecting and identifying all the molecules we come into contact with.

In scientific terms, the collection of all the molecules we are exposed to is called the “exposome chemical space” and it is at the heart of Samanipur's scientific endeavours. It is his mission to explore this vast molecular space and map even its most “remote” corners. He is driven by curiosity, but more than anything, by necessity.

Direct or indirect exposure to a myriad of mostly unknown chemicals poses a significant threat to human health: for example, it is estimated that 16% of premature deaths worldwide are pollution-related.

The environment is also suffering, including the loss of biodiversity, and Samanipour says it could be argued that humanity has exceeded its safe operating envelope when it comes to introducing man-made chemicals into the planetary system.

Current approaches are reactive in nature

“It's rather frustrating how little we know about this,” he said. “We know so little about the chemicals already in use, and yet there's no way we can keep up with the new chemicals that are being produced at an unprecedented pace.”

In previous studies, he estimated that less than 2 percent of the chemicals we're exposed to have been identified.

“The way society approaches this problem is essentially passive and at best reactive. It is only once we observe some effect of exposure to a chemical that we feel the urge to analyse it. We try to determine the presence of the chemical, its effects on the environment and human health, and by what mechanism it causes harm.”

“This has caused a lot of problems, the most recent being the crisis with PFAS chemicals, but we've also seen big problems with flame retardants, PCBs, CFCs and others,” he added.

Furthermore, regulatory actions are primarily targeted at chemicals with highly specific molecular structures that are produced in large quantities.

“There are countless chemicals out there that we don't know much about, and they're not just man-made,” Samanipour says. “Nature also produces chemicals that can harm us, either through completely natural synthetic pathways or by conversion of man-made chemicals.”

The latter category in particular has been systematically overlooked, according to Samanipour: “Traditional methods only catalog part of the exposome, often missing transformation products and leading to uncertain results.”

A data-driven approach is needed

The JACS Au paper provides a thorough review of the latest efforts in mapping the exposome chemical space and discusses their results. The main bottleneck is that traditional chemical analysis is biased towards known or proposed structures, which are key to interpreting data obtained by analytical methods such as chromatography and mass spectrometry (GC/LC-HRMS), thus overlooking more “unexpected” chemicals. This bias is circumvented by so-called non-targeted analysis (NTA), but results are still limited.

Over the past five years, 1,600 chemicals have been identified, with approximately 700 new chemicals being introduced each year into the U.S. market alone.

“Given the potential transformation products of these novel chemicals, we are forced to conclude that NTA research is moving too slowly to keep up,” says Samanipour. “As things stand, our chemical exposome will remain unknown.”

The paper lists these and many other bottlenecks in current analytical science and suggests ways to improve results. In particular, Samanipour argues that the use of machine learning and artificial intelligence will really move the field forward.

“A data-driven approach is needed from several directions,” he said. “First, data mining efforts need to be strengthened to extract information from existing chemical databases. Already documented structure-exposure-effect relationships of identified chemicals can lead to new insights. For example, they could help predict the health effects of related chemicals that have not yet been identified.”

“Second, retrospective analysis should be performed on existing analytical data obtained with established methods to expand the identified chemical space. We will certainly find molecules that have been overlooked so far. And third, we can use AI to work on understanding the structure and scope of the exposome chemical space.”

Of course, Samanipour knows that all this is a very complex, even daunting problem. But like real-life space explorers, he's an astronaut in molecular space and won't be put off by the complexity. “We have to work hard to tackle this problem. I have no illusions that in my career as a scientist I will be able to fully unravel the chemical space of the exposome. But it is essential that we face that complexity, discuss it and take the first steps to understand it,” he adds.

For more information:
Saer Samanipour et al. “Exploring the chemical space of the exposome: how far have we gone?” Jack's O (2024). DOI: 10.1021/jacsau.4c00220

Provided by University of Amsterdam

Quote: Only through machine learning and AI will we be able to understand all the chemicals around us, say researchers (July 1, 2024) Retrieved July 6, 2024 from https://phys.org/news/2024-07-machine-ai-chemicals.html

This document is subject to copyright. It may not be reproduced without written permission, except for fair dealing for the purposes of personal study or research. The content is provided for informational purposes only.





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *