AI tools predict functions of unknown proteins

A new artificial intelligence (AI) tool that draws logical inferences about the function of unknown proteins promises to help scientists uncover the inner workings of cells.

Developed by KAUST bioinformatics researcher Maxat Kulmanov and colleagues, the tool outperforms existing analytical methods for predicting protein function and can even analyze proteins for which there is no clear match in existing datasets. Masu.

This research nature machine intelligence.

The model, called DeepGO-SE, leverages large-scale language models similar to those used in generative AI tools such as Chat-GPT. Logical implications are then used to draw meaningful conclusions about molecular function based on general biological principles about how proteins work.

This is essentially building models of parts of the world (in this case, protein functions) and inferring the most plausible scenarios based on common sense and reasoning about what should happen in these world models. allows the computer to process the results logically.

“This method has many applications, especially for processing data and hypotheses generated by neural networks and other machine learning models,” said Robert Hoerndorff, head of the KAUST BioOntology Research Group, who oversaw the study. It’s useful when you need to make inferences.” ”

Kurmanov and Hoendorf, in collaboration with KAUST’s Stephan Arold and researchers at the Swiss Institute for Bioinformatics, evaluated the model’s ability to decipher the function of proteins whose roles in the body are unknown.

This tool successfully exploited data on the poorly understood protein’s amino acid sequence and known interactions with other proteins to accurately predict its molecular function. The model was so accurate that DeepGO-SE was ranked in the top 20 of more than 1,600 algorithms in the International Competition of Function Prediction Tools.

The KAUST team is currently using this tool to investigate the function of a mysterious protein discovered in plants that thrive in the extreme environments of Saudi Arabia’s deserts. They hope their findings will help identify novel proteins for biotechnological applications and hope other researchers will adopt this tool.

“DeepGO-SE’s uncharacterized protein analysis capabilities can facilitate tasks such as drug discovery, metabolic pathway analysis, disease association, protein engineering, and screening for specific proteins of interest,” said Kulmanov. explains.