Science1 publisher2 min readPublished
Murdoch review argues the enzymes to break down pollutants are already in the databases
Joseph Boctor of Murdoch University argues that machine learning should mine the millions of enzymes already in databases to break down pollutants. He frames it as a quicker response to contaminants that are already in the environment and causing harm.
The Scientist · Science desk
What happened
- The work comes out of Murdoch University's Bioplastics Innovation Hub, which pairs machine learning with biochemistry to pick out enzymes that can degrade pollutants.
- The team's next task is testing, validating and scaling the enzymes the search turns up, so that they can be used for actual bioremediation.
- A separate review Boctor published last year found agricultural soils holding roughly 23 times more microplastics than the oceans.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- capability Machine learning makes a catalogue too large to read by hand searchable, so a backlog of untested enzymes becomes something a lab can actually prioritise.
- decision Taken up, the approach would steer bioremediation effort toward finding and validating enzymes evolution already made and away from designing proteins to order.
- constraint Because the piece is a review setting a direction, it reports no hit rate or validated enzyme, so how well database mining works in practice is still an open question.
What Boctor published in Nature Reviews Earth & Environment is a review article [3]. His thesis is that the right enzyme for a given pollutant has probably already evolved somewhere and is waiting to be found [4]. Sorting through that much data by hand was never realistic, and machine learning is what makes the catalogue searchable [15]. The pipelines he describes learn from enzymes that have already been characterized, then predict how an enzyme's structure might latch onto a target pollutant and break it apart [7].
He is pointed about the alternative. "I strongly advocate that overengineering enzymes is a bad starting point that overlooks millions of years of evolution that have already produced lots of potential solutions to these contaminants," Boctor said [5]. "We need to look in the right place, and by leveraging machine learning tools, we are able to mine through millions of pieces of unexplored biological data to find the right candidate for the relevant task" [6].
He connects the search to health. "PFAS, microplastics and other persistent pollutants are not only industrially favorable but biologically active," he said [9]. "They trick our bodies by mimicking our own hormones and causing documented disruptions to health" [10]. In a review last year, Boctor reported microplastics and nanoplastics in lettuce, wheat and carrot crops [12], along with soil additives such as phthalates, linked to reproductive issues, and PBDEs, neurotoxic flame retardants linked to neurodegenerative disease, stroke, heart attack and early death [13].
What to watch
- Whether the Bioplastics Innovation Hub publishes validated enzymes and a success rate from its database searches.
- Whether any identified enzyme scales from lab testing to field cleanup of PFAS or microplastics.
- How other researchers respond to the claim that discovery should come before protein engineering.