CGRclust: Chaos Game Representation for twin contrastive clustering of unlabelled DNA sequences
2024

CGRclust: A New Method for Clustering DNA Sequences

Sample size: 2688 publication 10 minutes Evidence: high

Author Information

Author(s): Alipour Fatemeh, Hill Kathleen A., Kari Lila

Primary Institution: University of Waterloo

Hypothesis

Can unsupervised learning methods effectively cluster unlabelled DNA sequences without requiring alignment or labels?

Conclusion

CGRclust is a novel, scalable, alignment-free DNA sequence clustering method that achieves high accuracy across diverse datasets.

Supporting Evidence

  • CGRclust achieved over 80% accuracy in more than 90% of the datasets analyzed.
  • It outperformed existing methods like DeLUCS and MeShClust v3.0 in clustering accuracy.
  • CGRclust effectively handled unbalanced datasets, achieving high accuracy even in challenging scenarios.
  • The method demonstrated robustness across various genomic datasets, including mitochondrial and viral genomes.

Takeaway

CGRclust is a new way to group DNA sequences without needing to label them first, and it works really well on different types of DNA.

Methodology

CGRclust uses Chaos Game Representations and convolutional neural networks for unsupervised clustering of DNA sequences.

Limitations

The method may require fine-tuning of hyperparameters for different datasets, which can complicate its application.

Digital Object Identifier (DOI)

10.1186/s12864-024-11135-y

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication