Comparative Analysis of Feature Representation in DNA Sequence Similarity Using Alignment-Free Approaches

Abstract

Numerous methodologies have been proposed for DNA sequence similarity analysis, both Alignment-based and Alignment-free analysis. Though Alignment based methods give accurate results, in the last few decades, AF methods have proven to come close to being as effective as AB methods while overcoming a lot of limitations. Our research revealed which methods provide greater performance independently and which ones with different combination of possible methods. The feature representation has to be information lossless. Different representation methods are highlighted and a second step of feature generation has been applied if representation does not provide required features for sequence comparison. We find that the methods involving k-mer analysis achieve higher accuracy than other approaches. k-mer analysis combined with matrix reduction yielded the highest accuracy, maintaining the lowest RF score among the different techniques