New machine learning methods can better identify enzymatic metals in proteins

MAHOMES does not target a wide range of receptors, but distinguishes enzymes and non-enzymatic metals in proteins, with an accuracy rate of 92.

Corresponding author Joanna Slusky, associate professor of molecular biology and computational biology at the University of Kansas, said: "Enzymes are super interesting proteins.

Slusky and her lab's graduate collaborators, Ryan Feehan (a fan of the Chiefs, named MAHOMES) and Meghan Franklin of the University of Kansas Center for Computational Biology, tried to use computers to distinguish metalloproteins (without chemical reactions) from metalloenzymes.

The problem is that metalloproteins and metalloenzymes are the same in many ways

"People don't know exactly how enzymes work," Slusky said

As an undergraduate at the University of Kansas, co-lead author Feehan began to compile the world's largest structural data set of enzyme and non-enzymatic metalloprotein sites-this work continued into his career as a graduate student

Slusky said: "Structural data is difficult to obtain

Feehan found thousands of unique active and inactive metal binding sites, and then tested machine learning methods to distinguish these two sites

Slusky said this approach may be an important step in making enzymes more useful in the production of life-saving drug therapies and many other industrial processes

"I hope it will change the composition in general," she said

Slusky said that machine learning research will continue along three lines

"First, we are trying to make machine learning methods work better," she said

DOI

10.