kmer2vec: A Novel Method for Comparing DNA Sequences by word2vec Embedding.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Evolutionary Biology

Background:

• Traditional multiple sequence alignment (MSA) is computationally intensive for large datasets.
• Existing k-mer alignment-free methods may not fully capture sequence contextual structures.
• Genomic sequence comparison is crucial for evolutionary and phylogenetic analysis.

Purpose of the Study:

• To introduce kmer2vec, a novel k-mer contextual alignment-free method for DNA sequence comparison.
• To leverage natural language processing techniques (word2vec) for semantic embedding of k-mers.
• To improve the accuracy and speed of phylogenetic analysis, especially for large genomes.

Main Methods:

• Semantically embed sequence k-mers into word2vec vectors, capturing contextual relationships.
• Represent DNA/RNA sequences as points in a high-dimensional word2vec space for comparison.
• Optimize word2vec parameters and validate the method on large coronavirus and bacterial genomes.

Main Results:

• kmer2vec demonstrates effectiveness in phylogenetic tree construction and species clustering.
• The method significantly outperforms MSA in speed and conventional k-mer methods in accuracy.
• Accurate phylogenetic relationships are derived, enabling analysis of large-scale genomic data.

Conclusions:

• kmer2vec offers a computationally efficient and accurate approach for genomic sequence comparison.
• The method provides new insights into phylogeny and evolution, facilitating large genome analysis.
• Potential for rapid SARS-CoV-2 typing by combining kmer2vec with clustering techniques.