Artificial neural network method for discriminating coding regions of eukaryotic genes

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Eukaryotic gene structure involves complex regulatory elements.
• Accurate identification of coding regions is crucial for gene function analysis.
• Distinguishing true coding sequences from non-coding regions presents a computational challenge.

Purpose of the Study:

• To apply artificial neural networks (ANNs) for discriminating eukaryotic gene coding systems.
• To develop models for identifying key genetic features like promoter regions and splice sites.
• To assess the efficacy of ANNs in selecting correct coding regions.

Main Methods:

• Utilized over 300 eukaryotic genes from eight diverse species (human, mouse, rat, horse, ox, sheep, soybean, rabbit).
• Developed distinct ANNs models focusing on promoter regions, poly(A) signals, intron splice sites, and noose structures.
• Employed a fixed coding length parameter for discrimination tasks.

Main Results:

• ANNs demonstrated high accuracy in discriminating coding regions across various eukaryotic organisms.
• The models successfully identified correct coding sequences from a large set of potential solutions.
• Performance was consistent when the coding length was precisely defined.

Conclusions:

• Artificial neural networks provide a robust computational tool for eukaryotic gene identification.
• This approach enhances the accuracy of gene structure analysis and functional annotation.
• The method offers a reliable solution for resolving ambiguities in coding region determination.