Phylogenetic analysis of DNA sequences or genomes by Fourier transform

Phylogenetic analysis of DNA sequences or genomes by Fourier transform performs phylogenetic analysis of DNA sequences and genomes using a Discrete Fourier Transform (DFT)-based alignment-free method to compare sequences, quantify dissimilarity, and support phylogenetic tree construction while accommodating genomic rearrangements and length variation.

Key Features:

• Alignment-Free Methodology: Employs an alignment-free technique that leverages the full informational content of DNA sequences without relying on multiple sequence alignment, enabling analysis of rearranged sequences.
• Two-Dimensional Numerical Mapping: Maps DNA sequences into two-dimensional (2D) numerical representations that incorporate nucleotide composition to reduce bias in distance measures.
• Discrete Fourier Transform Application: Applies the Discrete Fourier Transform (DFT) to 2D numerical sequences to generate Fourier power spectra for sequence comparison.
• Improved Even Scaling Algorithm: Uses an improved even scaling algorithm to extend shorter DFT power spectra to the length of the longest sequence so all spectra share the same Fourier frequency dimensionality.
• Euclidean Distance as Dissimilarity Metric: Computes Euclidean distances between full Fourier power spectra to serve as dissimilarity metrics for sequence comparison.
• High Computational Efficiency and Accuracy: Achieves increased computational performance and accuracy, suitable for analysis of individual genes and large whole bacterial genomes.

Scientific Applications:

• Phylogenetic tree construction: Constructs phylogenetic trees from hierarchical clustering of Euclidean distances between Fourier power spectra.
• Comparison of rearranged and length-variable sequences: Enables reliable comparison of sequences with genomic rearrangements and differing lengths without multiple sequence alignment.
• Validation across datasets: Has been validated on simulated and real datasets and demonstrated efficacy across diverse DNA sequence length ranges.

Methodology:

DNA sequences are mapped into 2D numerical representations; DFT is applied to obtain Fourier power spectra; shorter spectra are extended using an improved even scaling algorithm; Euclidean distances between full power spectra are computed and used for hierarchical clustering to produce phylogenetic trees.