Classification of coding and non-coding regions in eukaryotic gene sequences using an adaptive anti-notch filter...

Area of Science:

• Genomics
• Bioinformatics
• Signal Processing

Background:

• Accurate classification of coding regions in gene sequences is vital for understanding protein formation and biological functions.
• Identifying coding regions is challenging due to their disparate distribution and low density in eukaryotic genomes.

Purpose of the Study:

• To propose a data-independent and window length-independent genomic signal processing technique for predicting coding and non-coding regions.
• To leverage the inherent period-3 characteristic of exons for improved classification accuracy.

Main Methods:

• A finite impulse response-based anti-notch filter (ANF) was designed using a hybrid Salp-Swarm-Whale optimization algorithm to capture the period-3 frequency.
• Gene sequences were converted to binary data using Voss representation and processed by the ANF.
• Low-pass filtering and power spectral density estimation were employed to identify period-3 peaks indicative of coding regions.

Main Results:

• The proposed method demonstrated sequence length independence, performing effectively on both short (∼100 bp) and long (∼1,000,000 bp) sequences.
• Evaluation on a benchmark sequence (F56F11.4a) yielded an AUC of 0.98 and an accuracy of 0.954.
• Testing on a diverse dataset (MOD191) of 191 sequences from various organisms resulted in an AUC of 0.91.

Conclusions:

• The developed genomic signal processing technique effectively identifies coding and non-coding regions by exploiting the period-3 characteristic of exons.
• The method's robustness across different sequence lengths and diverse organisms highlights its potential for broad genomic analysis.
• This approach offers a promising tool for accurate gene sequence classification in bioinformatics and genomics research.