Millions of enzymes present in scientific databases may hide clues to the degradation of plastics and other persistent pollutants. Australian researchers are turning to artificial intelligence tools to identify, among this vast amount of information, the molecules that may have the capacity to act on different contaminants.
The work is being developed at the Bioplastics Innovation Hub, a collaboration between Murdoch University and CSIRO, and combines machine learning with biochemistry to search for enzyme candidates for use in bioremediation processes.
In a review article published in Nature Reviews: Earth & Environment, researcher Joseph Boctor argues that part of the answers may already exist in nature and be recorded in databases, but have not yet been identified as a potential tool to combat certain pollutants.
“I argue that starting by excessively modifying enzymes is a bad approach, because it ignores millions of years of evolution that have already produced many potential solutions for these contaminants,” he says.
Searching among millions of possibilities
Machine learning tools enable analyzing large amounts of data about enzymes that have already been studied and using this information to predict how different structures may interact with certain pollutants.
Rather than trying to create a new enzyme from scratch, researchers thus search for molecules that already display characteristics potentially suitable for degrading a particular contaminant.
“We need to look in the right place and, by using machine learning tools, we can explore millions of biological data points not yet studied to find the right candidate for the task,” Boctor explains.
The process is still at the identification and validation stage. After the most promising enzymes are found, the team is testing their activity, confirming their function, and evaluating the possibility of using them at a larger scale.
From soil to food
Among the contaminants driving this investigation are microplastics and PFAS, a group of persistent chemicals often described as “forever chemicals” due to their high resistance to environmental degradation.
The researcher also notes the growing presence of different contaminants in agricultural soils. A review previously conducted by Boctor estimated that the agricultural soils analyzed contained microplastic concentrations far higher than those found in the oceans.
There are also studies that detected microplastics and nanoplastics in crops such as lettuce, wheat and carrots. Other contaminants found in soils include phthalates and PBDEs, substances used, among other applications, as plasticizers or flame retardants.
The researchers’ concern relates to the possibility that these compounds remain in the environment and enter different exposure pathways, including through food production.
A tool, not an immediate solution
For the Australian team, the use of artificial intelligence can accelerate a step that, otherwise, would require manual analysis of an amount of information virtually impossible to process.
The technology allows reducing the universe of potential candidates, but identifying an enzyme in a database does not mean that it can be immediately used to remove a pollutant from the environment.
The selected molecules still have to be tested in the laboratory, evaluated for their efficacy and safety, and studied under conditions that reproduce the environments where they may be applied.
The ultimate goal is to develop bioremediation tools capable of degrading contaminants already in the environment, while alternatives to reduce the use and production of persistent materials and substances continue to be sought.
