iDeLUCS

iDeLUCS performs unsupervised alignment-free clustering of DNA sequences by using deep learning to extract genomic signatures for downstream clustering.

Key Features:

• Alignment-Free Clustering: Clusters DNA sequences without requiring sequence alignment or taxonomic identifiers.
• Deep Learning Integration: Trains deep learning models to detect genomic signatures that serve as features for clustering.
• Hyper-parameter Tuning: Supports adjustment of training hyper-parameters for the underlying deep learning models.
• Scalability and Versatility: Validated on datasets from Animalia, Protista, Fungi, Bacteria, Archaea, viral genomes, simulated metagenomic reads, and synthetic DNA sequences.
• Performance Evaluation: Evaluated against k-means++, GMM, MeShClust v3.0, and DeLUCS, reporting approximately 20% higher accuracy versus classical algorithms and ~12% versus specialized tools on real DNA sequences.

Scientific Applications:

• Unsupervised Genomic Clustering: Grouping large-scale DNA sequence datasets without alignment or prior taxonomic labels.
• Evolutionary and Comparative Genomics: Detection of genomic signatures to explore evolutionary relationships and novel genetic patterns.
• Metagenomics: Clustering of simulated metagenomic reads to support community-level sequence analyses.
• Viral and Synthetic Sequence Analysis: Clustering and analysis of viral genomes and synthetic DNA sequences.

Methodology:

Trains deep learning models to extract genomic signatures from DNA sequences and uses those signatures as features for alignment-free clustering.