Scientists from King Abdullah University of Science and Technology (KAUST) in Saudi Arabia have developed DeepGO-SE, an artificial intelligence (AI) tool to study unknown properties of proteins. The study was published in the scientific journal magazine Nature Machine Intelligence (NMI).
DeepGO-SE leverages large language models such as ChatGPT. Using algorithms, a neural network can reveal the molecular functions of biological compounds.
This approach gives computers the ability to logically process results by creating models of protein properties not yet understood by scientists.
“This method has many applications, especially when it is necessary to draw conclusions from data and hypotheses generated by a neural network or other machine learning model,” said Robert Hendorf, head of the bioontology research group.
KAUST is using DeepGO-SE to investigate the functions of proteins found in plants thriving in the extreme desert conditions of Saudi Arabia.
DeepGO-SE was successfully used to decipher amino acid sequence data of a poorly understood protein and its known interactions with other proteins. The model accurately predicted the functions of the biological component. Based on the test results, DeepGO-SE ranked among the top 20 algorithms among more than 1,600 artificial intelligence tools for scientific prediction.
In the future, AI can be used for drug discovery, protein engineering, metabolic pathway analysis, and other research.
Previous scientists warned About the risk of humanity becoming stupid due to artificial intelligence.