Machine learning classifiers predict key genomic and evolutionary traits across the kingdoms of life

Area of Science:

• Genomics
• Bioinformatics
• Evolutionary Biology

Background:

• Codon usage bias (CUB) is a non-random pattern in DNA sequences.
• CUB varies across organisms and can reflect genomic and evolutionary pressures.
• Understanding CUB can provide insights into an organism's biology.

Purpose of the Study:

• To investigate codon usage bias as a predictor and classifier of genomic and evolutionary traits.
• To develop and apply AI/machine learning models for analyzing CUB.
• To assess the utility of CUB in phylogenetic classification and open reading frame detection.

Main Methods:

• Secondary analysis of existing genetic datasets.
• Development of AI/machine learning models trained on codon usage patterns from nearly 13,000 organisms.
• Application of a custom feature ranking ensemble for identifying influential codons.
• Statistical modeling for open reading frame (ORF) assessment based on codon usage distribution.

Main Results:

• AI/machine learning models accurately predicted the organelle of origin and taxonomic identity of nucleotide samples.
• Identification of key codons crucial for accurate phylogenetic prediction.
• Demonstration of CUB-based classification framework for ORF detection.
• Publicly available dataset and analyses for reproducibility.

Conclusions:

• Codon usage patterns serve as robust classifiers for taxonomic and phylogenetic features.
• The genetic code's structure can be leveraged for biological classification tasks.
• CUB analysis offers a powerful tool for genomic and evolutionary studies, including ORF identification.