Enhanced Viral Genome Classification Using Large Language Models

Strategic Insights -

Researchers from the University of Technology and Applied Sciences have developed a new approach to classify viral genomes using large language models. Their study explores the application of natural language processing (NLP) techniques to classify genomic sequences, with a focus on utilizing fine-tuned large language models (LLMs) such as DNABERT, DNAGPT, and GENA LM. The research demonstrates the effectiveness of these models in achieving high accuracy in viral genome classification, with DNAGPT achieving an accuracy of 96%.

Key Takeaways:

  • The study highlights the importance of classifying genomic sequences in the field of virology, with a vast repository of genomic sequences from various species being available.
  • The researchers utilized traditional algorithms such as Random Forest (RF), K-nearest neighbors (KNNs), Decision Tree (DT), and Naive Bayes (NB) for genome sequence classification, in addition to deep learning models.
  • The study demonstrates the effectiveness of fine-tuned large language models (LLMs) such as DNABERT, DNAGPT, and GENA LM in achieving high accuracy in viral genome classification.
  • DNAGPT achieved an accuracy of 96% in classifying viral genomes, exceeding the performance of state-of-the-art machine learning and deep learning models.
  • The study concludes that NLP techniques can be used to classify genomic sequences, with a focus on utilizing advanced LLMs.
  • The researchers obtained a vast repository of genomic sequences from various species, including humans, animals, plants, bacteria, and viruses, which tend to mutate and form new variants or strains.

Statistics:

  • DNAGPT achieved an accuracy of 96% in classifying viral genomes.
  • The study utilized traditional algorithms such as Random Forest (RF), K-nearest neighbors (KNNs), Decision Tree (DT), and Naive Bayes (NB) for genome sequence classification.
  • The researchers obtained a vast repository of genomic sequences from various species, including humans, animals, plants, bacteria, and viruses.

Sources:

  • Algorithms, 2025,18(6):302. (Algorithms - http://www.mdpi.com/journal/algorithms). The publisher for Algorithms is MDPI AG.
  • NewsRx. University of Technology and Applied Sciences Researchers Reveal New Findings on Machine Learning (Enhanced Viral Genome Classification Using Large Language Models). Life Science Weekly. July 8, 2025; p 7716.