ChemGenome2.1: An Ab Initio Gene Prediction Software.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Gene prediction is crucial in molecular biology, especially with large-scale genome sequencing.
• Accurate gene identification is essential for understanding genome function and evolution.

Purpose of the Study:

• To present the methodology of ChemGenome2.1 (CG2.1), an advanced ab initio model for gene prediction in prokaryotic genomes.
• To evaluate the potential applicability of CG2.1 for gene finding in lower eukaryotes.

Main Methods:

• Utilizing a three-dimensional vector based on codon physicochemical properties: base pairing energy, stacking energy, and protein-nucleic acid interaction propensity.
• Employing a filter based on protein sequence stereochemical properties to reduce false positives.
• Final screening using amino acid frequencies and deviations from Swissprot values at monomer and tripeptide levels (Z-score).

Main Results:

• The CG2.1 model effectively distinguishes gene from non-genic regions in prokaryotes based on codon vectors.
• The stereochemical filter successfully reduces false positive gene predictions.
• Initial evaluations suggest CG2.1's physicochemical approach holds potential for eukaryotic gene finding.

Conclusions:

• The physicochemical model of codons provides a viable strategy for gene finding in prokaryotes.
• CG2.1 demonstrates potential for application in lower eukaryotes, indicating broader applicability of the physicochemical approach.
• Further improvements are necessary for optimal gene prediction in eukaryotes.