Sign in Subscribe
Machine learning

Researchers at University of Montpellier Target Science with Innovative Approach

Strategic Insights

— 1 min read

Researchers at the University of Montpellier have made a breakthrough in understanding species evolution and enabling functional annotation transfer by inferring protein homology from sequence information. They represented proteins with a biologically oriented large language model and applied k-means clustering to the embedded data to extract homology relationships. This approach lacks the sensitivity of other tools but obtains better precision for detecting n:m orthologs and reconstructs full orthologous groups from scratch.

Key Takeaways:

  • Researchers developed a machine learning approach to infer protein homology from sequence information, using a biologically oriented large language model and k-means clustering.
  • The approach obtained better precision for detecting n:m orthologs compared to other tools, despite lacking sensitivity.
  • The model was able to reconstruct full orthologous groups from scratch, highlighting its potential for analyzing protein data.
  • The research used k-means clustering to extract homology relationships from proteins embedded with a large language model.
  • The study concluded that the approach can be used in combination with clustering algorithms for protein data analysis.
  • The datasets used in the study are available on OrthoMCL-DB, and the source code can be accessed on GitHub and Zenodo.
  • The research has been published in Bioinformatics and has been peer-reviewed.
  • The study has the potential to improve our understanding of species evolution and enable functional annotation transfer.

Statistics:

  • The approach obtained a precision rate of [ omitted from the text ] for detecting n:m orthologs.
  • The model was able to reconstruct [ omitted from the text ] full orthologous groups from scratch.
  • The research used a biologically oriented large language model to represent proteins.
  • The k-means clustering algorithm was applied to the embedded data to extract homology relationships.
  • The study has been published in Bioinformatics and has been peer-reviewed.
  • The research has the potential to provide a better understanding of species evolution and enable functional annotation transfer.

Sources:

  • Exploring homology detection via k-means clustering of proteins embedded with a large language model. Bioinformatics, 2025.
  • NewsRx. Researchers at University of Montpellier Target Science (Exploring homology detection via k-means clustering of proteins embedded with a large language model). Science Letter. September 19, 2025; p 3671.

Read more