Articles
THE ABILITY FOR FOURIER TRANSFORM INFRARED SPECTROSCOPY TO CLASSIFY PERSIMMON GENOTYPES BY EPICUTICULAR LEAF WAXES
Article number
601_7
Pages
65 – 69
Language
English
Abstract
The ability of Fourier Transform Infrared (FTIR) spectroscopic measurements of epicuticular leaf waxes to classify five non-astringent persimmon cultivars (Diospyros kaki L. cv.
Fuyu, cv.
Matsumoto wase Fuyu, cv.
Hanna Fuyu, cv.
Oku Gosho and cv.
II-IQ-12) was investigated.
Principal component analysis was used to reduce the FTIR spectral data to 20 variables.
Machine learning models classified the cultivars into five groups.
Three of the cultivars, Oku Gosho, Hanna Fuyu and II-IQ-12 were readily distinguished from each other and from Fuyu and Matsumoto wase Fuyu.
However, the leaf wax data was unable to distinguished between two closely related persimmon genotypes, Fuyu and the early maturing L1 mutation of Fuyu, Matsumoto wase Fuyu.
Here, epicuticular wax biosynthesis appears unaffected by the LI mutation found in Matsumoto wase Fuyu.
When the classification was based on four groups, where Fuyu and Matsumoto wase Fuyu samples were combined, the best performing machine learning model correctly classified 98.7% of the samples.
Fuyu, cv.
Matsumoto wase Fuyu, cv.
Hanna Fuyu, cv.
Oku Gosho and cv.
II-IQ-12) was investigated.
Principal component analysis was used to reduce the FTIR spectral data to 20 variables.
Machine learning models classified the cultivars into five groups.
Three of the cultivars, Oku Gosho, Hanna Fuyu and II-IQ-12 were readily distinguished from each other and from Fuyu and Matsumoto wase Fuyu.
However, the leaf wax data was unable to distinguished between two closely related persimmon genotypes, Fuyu and the early maturing L1 mutation of Fuyu, Matsumoto wase Fuyu.
Here, epicuticular wax biosynthesis appears unaffected by the LI mutation found in Matsumoto wase Fuyu.
When the classification was based on four groups, where Fuyu and Matsumoto wase Fuyu samples were combined, the best performing machine learning model correctly classified 98.7% of the samples.
Publication
Authors
A.D. Mowat, G. Holmes
Keywords
Online Articles (34)