Objective To provide theoretical guidance for the phylogenetics, genetic diversity, and species identification of Dioscorea by integrating chloroplast genomics, codon usage bias, DNA barcoding, and machine learning technologies. Methods The chloroplast gene structures of 11 plant species were compared via IRscope and a phylogenetic tree was constructed. Codon usage bias was analyzed using tools such as CodonW and CUSP. Dioscorea materials were collected, and DNA barcode sequences were amplified and sequenced. Molecular identification was performed based on the maturase K gene (matK), photosystem b a protein gene-transfer RNA-Histidine intergenic spacer (psbA-trnH), and ribulose-bisphosphate carboxylase/oxygenase large subunit gene (rbcL) barcodes using machine learning algorithms. Results The chloroplast genomes of Dioscorea were found to be conserved and stable. Codon usage showed a significant A/T bias, with the third codon position favoring A/U endings. Natural selection was the primary factor influencing codon bias, and the common optimal codons identified were GGA and UCA. All three barcodes successfully discriminated species. Single-barcode identification using the BLOG algorithm achieved 100% success rate, while the SMO and NaïveBayes classifiers in WEKA demonstrated high identification accuracy. Conclusion The chloroplast genome of Dioscorea is conserved, and natural selection dominates its codon usage pattern. This study provides a basis and guidance for research on gene expression regulation, species identification, and resource conservation of Dioscorea.

Dioscorea L.  /  chloroplast genome  /  codon usage bias  /  DNA barcoding  /  machine learning  /  molecular identification