Prediction of Enzyme Functions Better Through Key Tools By AI

A novel artificial intelligence tool can help predict the functions of enzymes on the basis of their amino acid sequences, even when the enzymes are unstudied or not clearly understood.

As per the researchers, the AI tool called CLEAN outshines the leading top-notch tools in accuracy, dependency, and sensitivity. Apparently, a clearer understanding of enzymes as well as their functions will prove to be a big push for research as far as chemistry, genomics, medicine, pharma, industrial materials, etc. are concerned.

The study leader, Huimin Zhao, from the University of Illinois, states that just as the ChatGPT makes use of the data from written languages to create predictive text, they happen to be leveraging the protein language so as to predict their activity. He adds that almost all the researchers, when working on a novel protein sequence, intend to know without delay what the protein is up to. Besides, when it comes to making any applications, be it in medicine, biology, or industry, this tool will aid researchers in rapidly identifying the appropriate enzymes that are required for the synthesis of materials and chemicals.

The researchers are on the verge of publishing their assessment and, thereby, making CLEAN accessible online from March 31.

Due to the advances made in genomics, numerous enzymes have been pinpointed as well as sequenced, but the researchers have little or no data on what the enzymes are up to, commented Zhao, one of the members of the Carl R. Woese Institute at Illinois.

Apart from this, other computational tools intend to predict the enzyme functions. They look to assign an enzyme commission number, which happens to be an ID code that puts into focus what kind of reaction an enzyme can catalyse by way of comparing queried sequences to that of the enzyme catalogue as well as findings that are of similar sequences. That said, these tools do not work as well as they should with the less studied enzymes or even with enzymes that are known to perform multiple tasks.

Zhao adds that they aren’t the first when it comes to using the AI tools to forecast the enzyme commission numbers, but they are indeed the first to use the novel deep learning algorithm that happens to be working much better as compared to the AI tools used by others.

There is no guarantee that everybody’s product will be accurately predicted, but having said that, there is always the possibility of getting more accuracy compared to the other couple of methods.

The researchers have already verified the tool by way of an experiment using both in vitro and computational experiments. The inference was that not only was the tool able to predict the operations of the previously uncharacterized enzymes, but the correction of mislabelled enzymes by a leading software was done with either two or more functions.

Apparently, Zhao’s group is planning to make CLEAN accessible online so that other researchers who are looking to characterise an enzyme or gauge whether it could catalyse a desired location can do so. They hope that this tool is going to be helpful and used widely by the broad research community. The researchers just have to enter the sequence in a search box by way of a web interface and get the results.

As per Zhao, the group is planning to broaden the AI behind CLEAN so as to characterise certain other proteins as well, like the binding ones. The team is hopeful about further developing the machine learning algorithms so that the users search for a desired location and the AI pinpoints the proper enzyme for the job.

He further opines that there are many uncharacterized binding proteins, like the receptors, as well as transcription factors. They are looking to predict the functions of each and every protein so they are able to gauge all the proteins a cell has and thereby better study the entire cell for applications related to biomedical or biotech.