Different DNA Encoding Schemes for Filter Bank Based Tandem Repeat Detection

This paper tackles the challenge of detecting tandem repeats in DNA sequences, which play a key role in understanding genetic disorders, forensics, and parentage analysis. While current algorithms—both statistical and non-statistical—have proven effective, the computational complexity of integer encoding in the recently proposed Ramanujan Filter Bank (RFB) limits its practical usability. To address this, we introduce alternative encoding schemes, including 2-simplex, complex, and Voss mappings, designed to reduce computational costs. Additionally, we explore the same detection task using a filter bank based on Complex Conjugate Pair Sums (CCPSs), achieving detection accuracy comparable to that of the RFB. Through extensive numerical simulations on diverse gene sequences, we assess the performance of both filter banks in terms of F1-scores and computational complexity, showing that our proposed encoding schemes improve efficiency without compromising accuracy in tandem repeat detection across both filter bank frameworks.