A clustering package for nucleotide sequences using Laplacian Eigenmaps and Gaussian Mixture Model.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate clustering of nucleotide sequences is crucial for genomic analysis and understanding evolutionary relationships.
• Existing methods may require prior knowledge of cluster numbers or struggle with divergent sequences.

Purpose of the Study:

• To introduce a novel Python package for unsupervised DNA sequence clustering.
• To provide a method that accurately identifies clusters and their optimal number without prior specification.
• To offer a visualization tool for DNA sequence clusters.

Main Methods:

• Implementation of a Python package utilizing Laplacian eigenmap embedding and Gaussian Mixture Models.
• Input processing of nucleotide sequences for clustering.
• Generation of optimal cluster numbers and visualizations.

Main Results:

• The proposed method demonstrates superior accuracy in DNA sequence clustering compared to state-of-the-art tools like Cd-hit-est and BLASTClust.
• Clusters generated are consistent with phylogenetic analysis and NCBI taxonomy for a set of ND3 gene sequences.
• The approach effectively recovers clusters from synthesized data, outperforming existing software.

Conclusions:

• The new Python package provides an accurate and efficient solution for DNA sequence clustering.
• It offers advantages over existing methods by not requiring a priori knowledge of cluster numbers and handling divergent sequences effectively.
• The tool aids in genomic data analysis by providing accurate clustering and visualization consistent with evolutionary and taxonomic information.