Corehunter3: core collection inference with phenotypic data

Core collection inference

Brainstorming

During my adventure here in South Korea I was instructed to infer a core collection from a large phenotyping trials data set. After goggling I found an interesting paper from Tchokponhoué et al. (2020). The authors used corehunter3 package under R environment. So I decide to test it. First of all, let’s intall the package.

Installation

Step 1: Install java version 8 or more

Make sure to install Java Oracle JDK version 8 or more. I used a window OS. The CRAN github page recommended to download the java here. I download the Oracle Java JDK installer and installed.

Step2: Install rJava and corehunter

To install rJava and corehunter, do:

install.packages("corehunter")

install.packages("rJava")

It is important to check if the package are well installed. Thus do:

library(rJava)
library(corehunter)

Step3: Prepare the phenotype data set

Corehunter need a specific data format. The details of the specifications are well explained here.

The data that I will use in this tutorial look like this:

ID	Fruit_diam	Seed_wght	nr_flow_axila	Fruit_wght	Plant_hght	Fruit_length	Seed_nr	Corola_color	Anther_color	Corola_spot_color	Fruit_epid	Fruit_form	Flower_position	Corola_form
TYPE	R	R	R	R	R	R	R	NI	NI	NI	NI	NI	NI	NI
10	20	0.3	1	69.4	40.8	37.3	32.1	4	2	0	2	2	5	1
11	9.5	0.2	2	6.7	33.7	10.7	17.5	4	2	0	2	2	5	1
18	24.9	0.4	1.3	122.4	38.3	51	42.1	4	2	0	2	2	5	1
41	16.3	0.4	1.7	28	43.8	47.1	15.8	4	2	0	2	2	5	1
43	14.3	0.4	2	35.5	42.7	55.3	43.3	4	2	0	2	2	5	1
50	29.2	0.4	2	104.1	55.6	51.8	84.6	4	2	0	2	2	5	1
51	21.9	0.4	2	85.8	50.6	57.2	46.7	4	2	0	2	2	5	1
54	27	0.4	2	79.4	56.6	55	54.4	4	2	0	2	2	5	1
67	14.9	0.5	1.7	30.4	62.7	38.3	17	4	2	0	2	2	5	1
68	14.7	0.3	2	21.3	56.7	21.9	43.1	4	2	0	2	2	5	1
85	19.2	0.5	1.3	41.8	62.2	27.3	43.7	4	2	0	2	2	5	1
147	26.5	0.7	1	73.5	51.9	23.5	93.7	1	2	0	2	2	5	1
149	22.4	0.6	1	55.8	54	21.7	83.5	1	2	0	2	2	5	1
167	21.7	0.5	1	63.3	63.4	36.9	101.1	1	2	0	2	2	5	1
175	18.3	0.6	1	146.5	58.6	82.2	82.3	1	2	0	2	2	5	1
179	11.8	0.5	1	38.8	69.8	68.1	84.7	1	2	0	2	2	5	1
181	20	0.6	1	51.4	62.8	35.8	105.5	1	2	1	2	2	5	1
205	33	0.5	1.3	198.7	59.1	63.8	77.3	1	2	1	2	2	5	1
212	11.3	0.3	2	15.9	40.6	19.7	22.1	1	2	1	2	2	5	1
215	11.1	0.4	2	22.1	49.6	44	43.2	1	2	1	2	2	5	1
216	4.9	0.4	2	5.1	65.4	19.5	16.7	1	2	1	2	2	5	2
221	11.8	0.3	2	13.5	63.1	18.3	29.1	1	2	1	2	2	6	2
224	17.7	0.5	2.3	36.7	51	48.1	47.7	1	3	1	1	2	6	2
225	13.1	0.3	2	12.9	34.7	12.8	14.6	1	3	1	1	2	6	2
233	28	0.3	2.7	86.7	40.9	46.7	80.7	1	3	1	1	2	6	2
239	10.3	0.4	1.3	22.9	58	38.2	20.5	1	3	1	1	3	6	2
242	4.4	0.4	1.3	5.7	71.6	25.3	9.7	1	3	1	1	3	6	2
246	15.2	0.3	2	33.9	46.5	42.2	38.9	1	3	1	1	3	6	2
250	16.3	0.4	2	41.7	58.5	48	37.5	1	3	1	1	3	6	2
252	18.6	0.4	1.3	32.3	51.3	20.6	48.5	1	3	1	1	3	6	2
268	10.6	0.4	1.7	12.4	59.4	42.6	22.4	4	3	1	1	3	6	2
275	18.8	0.4	2	102.9	58.9	82.9	75.3	4	3	1	1	3	6	3
298	11.1	0.3	2	6.8	56.4	11	18.9	4	3	1	1	3	6	3
301	15.3	0.3	2	17.1	48	31.5	34.7	4	3	1	1	3	7	3
340	12.7	0.4	1	31.9	53.3	43.9	53.8	4	3	1	1	3	7	3
347	21.5	0.6	1	160.3	66.1	97.1	118.4	4	3	1	1	3	7	3
363	27.3	0.7	1	127.4	64.1	48.2	58	4	3	1	1	3	7	3
406	22.5	0.3	1.3	28.5	54.5	16.6	46.1	4	3	1	1	3	7	3
407	22.1	0.5	1.3	53.7	54	29.3	53.9	4	3	1	1	3	7	3

Real phenotypic trait data including 39 accessions for which seven quantitative and seven qualitative traits were recorded. Here, quantitative traits are ratio variables (R) encoded as double values (default for ratio variables) while qualitative traits are integer valued nominal variables (NI).

Solution

Let’s import this data in Rstudio

data_pheno = read.csv("phenotypic_data.csv", h = T, sep = ";")

An important step is to convert the csv file into corehunter format using the following code:

phenot <- phenotypes(pheno_data)

Now, let’s infer the core collection

core <- sampleCore(phenot)

To see the results, just call core

core

The results are presented like this:

$sel
[1] "1"  "12" "18" "23" "24" "32" "34" "37"

$EN
$EN$GD
[1] 0.6833333


attr(,"class")
[1] "chcore" "list"

The accession "1" "12" "18" "23" "24" "32" "34" "37" are the core collection inferred by corehunter

That’s it! The key point in this process is to properly arrange the data following the instructions well defined here.