Extraction of high quality k-words for alignment-free sequence comparison

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• The weighted Euclidean distance (D(2)) is a foundational alignment-free method for biological sequence comparison.
• D(2) relies on k-word counts, offering computational speed but traditionally processing all k-words.
• Variants of D(2) have been developed to enhance its applicability.

Purpose of the Study:

• To investigate if a reduced subset of k-words can maintain sequence comparison accuracy.
• To introduce a novel method for selecting informative k-words.
• To improve the efficiency of alignment-free sequence comparison techniques.

Main Methods:

• A term variance-based quality measure was developed to identify important k-words.
• The effectiveness of using selected k-word subsets was evaluated in phylogeny reconstruction.
• Exploratory analysis was conducted to determine optimal k-word values and subset impacts.

Main Results:

• Similar accuracy in sequence comparison was achieved using a selected subset of k-words compared to using all k-words.
• Up to 99% of k-words could be filtered out for specific datasets, leading to faster comparisons.
• The study identified optimal k-word subsets and analyzed their influence on results.

Conclusions:

• Selective k-word usage offers a viable strategy for accelerating biological sequence comparison.
• The proposed variance-based method effectively identifies key k-words for efficient analysis.
• This approach has significant implications for large-scale genomic and phylogenetic studies.