NetStart

NetStart predicts translation start sites in nucleotide sequences of vertebrates and Arabidopsis thaliana using artificial neural networks to assess context-dependent AUG initiation.

Key Features:

• Artificial Neural Networks (ANNs): NetStart employs artificial neural networks to analyze nucleotide sequences and predict which AUG triplet functions as the translation start codon.
• Predictive Accuracy: The trained neural networks correctly classify 88% of Arabidopsis AUG triplets and 85% of vertebrate AUG triplets.
• Contextual Analysis: NetStart integrates both local start codon context and global sequence information to inform predictions.
• Reading Frame Detection: Analysis of false predictions shows AUGs in the same reading frame as the actual start codon are more frequently selected, indicating implicit frame discrimination by the networks.
• Database Validation: NetStart identifies conflicts between its predictions and existing database annotations to flag potential annotation errors.

Scientific Applications:

• EST and genome analysis: Predicts translation initiation sites in EST (Expressed Sequence Tag) and genome data where the full mature mRNA sequence may be unknown.
• Gene expression and protein synthesis studies: Supports investigations into context-dependent translation initiation relevant to gene expression regulation and protein synthesis in eukaryotic organisms.
• Genomic database curation: Aids refinement of genomic databases by identifying possible annotation errors in reported start codons.

Methodology:

NetStart uses artificial neural networks trained on nucleotide sequences, integrates local start codon context and global sequence information, analyzes false predictions to assess reading-frame bias, and compares predictions with database annotations.