Genomic and subgenomic group discrimination between 100 Indian banana (Musa) accessions using ripe banana pulp multi-elemental fingerprints and chemometrics

• Published in: Journal of food composition and analysis 2024-07, Vol.131, p.106205, Article 106205
• Main Authors: Devarajan, Ramajayam, Dibakoane, Siphosethu R., Wokadala, Obiro Cuthbert, Meiring, Belinda, Mlambo, Victor, Kutu, Funso Raphael, Sibanyoni, July Johannes, Jayaraman, Jeyabaskaran Kandallu
• Format: Article
• Language: English
• Subjects: Banana Genome; bananas; Bananas (Musa); calibration; chemometrics; Chemometrics, Multi-elemental Fingerprints; discriminant analysis; food composition; genomics; India banana accessions; Musa; neural networks; principal component analysis; pulp; spectroscopy; Sub-genome groups; support vector machines
• ISSN: 0889-1575; 1096-0481
• DOI: 10.1016/j.jfca.2024.106205

Worldwide, there are over 1000 banana types which are classified in various subgenomic and genomic groups. Distinguishing between the banana types, their genomic and subgenomic groups has been a challenge due to different identities and nomenclature used in different regions of the world. The present study assessed the efficacy of multi-elemental fingerprinting combined with chemometrics to distinguish between genomic and sub-genomic groups within 100 Indian banana (Musa) accessions based on ripe banana pulp elemental concentrations. The concentrations of B, Ca, Fe, Mg, Mn K, Zn, Na, and P were analyzed using Inductively Coupled Plasma-Optical Emission Spectroscopy (ICP-OES). Multi-elemental fingerprints plus chemometrics were done using principal component analysis (PCA) then combined with linear discriminant analysis (PCA-LDA), support vector machine (PCA-SVM), and artificial neural network (PCA-ANN) for classification analysis with an 80:20 split between the calibration and verification sets (with total of 300 specimens). The PCA-SVM model was the most effective in classification when applied to the verification set subgenomic and genomic groups data, with accuracies of 83.7% and 100.0% respectively. These results demonstrated that ripe banana pulp multi-elemental fingerprints combined with chemometrics can discriminate between genomic and sub-genomic groups for Indian banana (Musa) accessions. •The multi-element concentrations of 100 Indian Musa accessions were determined.•Multi-elemental fingerprints were determined through principal component analysis (PCA).•PCA-SVM, PCA-ANN and PCA-LDA classification models were developed.•Verification set classification accuracy ranged from 83.7% to 100% with PCA-SVM.