phenoclass: Tools for setting objective quantitative levels of expression in characterizations in R.

This mini-package allows to define levels of expression for quantitative traits following the methodology described in the book “Harmonized methodology for the pomological characterization of apple (Malus x domestica Borkh.)” by Royo et al (2017). This method is based on the UPOV test guidelines, but it redefines rigorously every characteristic and level in order to minimize subjectivity, therefore allowing to compare easily descriptions made by different teams. The amount of levels for any characteristic depends on the variability in the values that can be found within the same accession and within accessions. For that reason, for that reason, the width of the intervals of the scale (called in the methodology as Discrimination Unit or DU) is greater than the difference that, on average, can be found between the largest and smallest measurable values within the same accession. As a consequence, the number of classes is not defined a priori, but depends on the value of the DU and the mean value of the characteristic in the accessions with extreme values in the collection to be classified.

---

Resources

• Installation
• 1. Required package
• 2. Example: Estimate the phenology of a peach cultivar

---

Installation

You can install phenoclass from GitHub with:

install.packages("devtools")
library(devtools)
devtools::install_github("Carm1r/phenoclass")

Menu

---

Using phenoclass

1. Required package

• tidyverse

install.packages("tidyverse")

library(phenoclass)
library(tidyverse)

Menu

---

2. Example. Definition of the classes for a character in pear flowers

This example shows how to use the functions clean_outliers and calc_du to define the number of phenotypic classes that can be observed without ambiguity using dummy characterization data. The example uses the dataset U35_Pe, included in the package, which contains data for the length of claw of petal in pear flowers, corresponding to descriptor U35 in TG/15/3 (UPOV, 2000). The dataset contains lengths for 52 accessions, ten observations per accession and year, evaluated for up to three years.

library(phenoclass)
library(tidyverse)
# Remove outliers using a thr_y=1.5 and thr_g =1.8
data(U35_PE)
Claw.clean <- clean_outliers(U35_PE, 1.5, 1.8)
# Define the central class and DU for the accessions
Claw.class <- calc_du(Claw.clean)

The procedure provides the following results:

• Nacc: 52
• Nacc: 43
• vm_g: 0.781
• VM_g: 2.864
• DU : 0.959
• P50 : 1.757
• Central class: 1.28 - 2.24

That is, data for 43 accessions has been used to define the scale, the remaining 9 accessions were found too variable and would distort the classification, providing a scale with intervals too wide to be of use. The accessions with the lowest (vm_g) and highest (VM_g) values, respectively. The discrimination unit is 0.959 mm, that is, accessions differing more than that value will be assigned to different levels. As the median value of the accesions for that characteristic is 1.757 mm, the central class is 1.28 - 2.24. Below the central class there is just one level, as 1.28-0.96 = 0.32, which is smaller than the minimum level observed. Above the central class, only one addtional level can be defined, as 2.24 + 0.96 = 3.20, wich is above the maximum value observed in the collection. Therefore, within the accessions included in the dataset, only three levels can be observed, as follows:

• Level 1: <1.28mm
• Level 2: 1.28mm - 2.24mm
• Level 3: >2.24

Menu

Licenses

The R/phenoclass package as a whole is distributed under GPL-3 (GNU General Public License version 3).

Acknowledgements

This research has been funded by project PID2019-108081RR-C22 (APPLECUT) funded by MCIN/ AEI /10.13039/501100011033

Author

Carlos Miranda