Researchers created DeepGO-SE, an AI tool excelling in predicting unknown proteins’ functions, a major leap in bioinformatics. Using large language models and logical entailment, it deduces molecular functions for proteins without database matches, offering a groundbreaking approach to cellular mechanisms. DeepGO-SE, ranking top 20 among 1,600 algorithms, excels in drug discovery, metabolic pathway analysis, and more. Its precision surpasses existing methods for characterizing proteins, even those uncharacterized before. Applying logical entailment and large language models, it infers functions based on biological principles and amino acid sequences. Developed by KAUST researcher Maxat Kulmanov, DeepGO-SE’s accuracy impressed in an international competition. Leveraging models similar to Chat-GPT, it draws meaningful conclusions about molecular functions, aiding tasks like drug discovery, metabolic analysis, and protein engineering. The KAUST team now explores proteins in extreme environments for biotechnological advancements.
Source KAUST
