Multi-character selection for identifying superior oil palm cultivar using MGIDI

Efficient selection methods are needed in oil palm breeding programs. Selection efficiency will increase if the selection is conducted on several characters simultaneously. However, the selection of segregating progeny is difficult because it is quantitative. MGIDI (Multi-trait Genotype Ideotype Distance Index) is formed based on Monte Carlo simulation to select superior progeny by considering several characters simultaneously. The aim of this study is to evaluate and identify planting materials with compact palm and good production traits using MGIDI. Data were obtained from 21 DxT palms planted at Dolok Sinumbah Plantation, PT. Nusantara IV Plantation, North Sumatra. The imbalanced North Carolina II mating design was employed to produce the materials for this study. The observed characters are the bunch analysis component and vegetative traits. Based on the factorial analysis (FA) results, these characters were grouped into three groups where FA1 contains ODM, OWM, HI, and DIAM. FA2 consists of BW and FB. FA3 consists of MF, OB, SF, KF, and LAI. Based on MGIDI, the selected superior progenies were DT11, DT6, and DT1. Aside from these genotypes DT15 was near the cut point, suggesting that it has an interesting feature. Positive selection gains (0.57-1.52) are obtained for the desired traits, and negative selection gains (-0.24 – (-0.88)) for undesirable traits. Visualization of the strength and weaknesses of each genotype can also assist breeders in decision-making to improve the progeny in the next breeding cycle.


BACKGROUND
Palm oil plays an important role in Indonesian foreign exchange, contributing US$ 35.79 billion in 2022 [1].Oil yield improvement is a fundamental goal in oil palm breeding.Oil palm trees can grow quite tall, selecting oil palm with low height increment and compactness can make harvesting more accessible and efficient.The process of enhancing and distributing superior oil palm genotype through genetic means is a recurring obstacle for oil palm breeding.This task typically necessitates a breeding cycle of 12 to 20 years, thereby posing a significant challenge [2].Thus, selecting the best breeds will produce superior cultivars, increasing palm oil productivity per hectare.
Oil palm cultivars are developed by utilizing a plant ideotype that encompasses a collection of favorable characteristics that are individually evaluated in the field.Nevertheless, identifying genotypes with exceptional performance in multiple traits concurrently is a challenging undertaking.The selection of economic-related-traits in segregating progeny is challenging because it is quantitative.In addition, if these characters are correlated with one another, choosing one character may result in alterations to other characters [3].An approach that can be used to select several characters is the index selection, which gives weight to the observed traits [4].However, the index selection coefficient estimation can be negatively affected by multicollinearity effects between the tested characters [5].The use of inefficient selection methods may hinder the breeder's capacity to identify the most optimal genotypes, particularly when realistic economic weightings are taken into consideration [6].
In order to address the issue of multicollinearity, a new multi-trait index called MGIDI was proposed by the authors of reference [7].This index was able to generate scenarios by varying genotypes, evaluated traits, and correlation structures of two parameters such as factorial analysis and genotype-ideotype distance.The weighting process in MGIDI involves multiplying each element in the original data with a weight vector in rescaled data.Several studies show MGIDI is able to identify superior genotypes on desirable traits [8] [9].In this paper, MGIDI is used to evaluate and identify planting materials with compact palm and good production traits.

Plant material and experimental
Data was collected from 21 DxT progeny tested derived from crossing between 7 dura dan 11 pisifera using the unbalanced North Carolina II mating design.The female parent in both groups originated from recombining the best Deli dura in the second cycle.The male parent consists of two groups: the recombination of pisifera La Me with SP540 population and the Binga population.The trial planted at Dolok Sinumbah, Afdeling III, PT.Perkebunan Nusantara IV in randomize complete block design (RCBD).

Assessed traits
The observations were conducted on plants with an age of 5 to 8 years.The bunch quality component consists of bunch weight (BW), fruit to bunch (FB), mesocarp to fruit (MF), oil to dry mesocarp (ODM), oil to wet mesocarp (OWM), oil to bunch (OB), kernel to fruit (KF), and shell to fruit (SF).Vegetative traits such as height increment (HI), trunk diameter (DIAM), rachis length (RL), and leaf area index (LAI) were also assessed to gain a comprehensive understanding of the plant's growth.

Defining ideal ideotype
Based on previous knowledge, an ideal oil palm ideotype should provide: (i) more erect canopies that can tolerate high-density planting with low height increment to improve harvesting efficiency.Thus, lower values for HI, RL, and LAI are desired [10] [11], (ii) high fruit yield and oil content, defined by higher values for BW, FB, MF, ODM, OWM, and OB [12], (iii) high lipid-rich fleshy mesocarp tissue of the oil palm.Thus, lower values for KF and SF are desired [13].High planting density and high oil to bunch ratio can contribute to the sustainability of oil palm plantations.

Statistical analysis
The computation was performed in the Rstudio using the 'metan package'.The calculation of MGIDI involved three primary stages: rescaling means, factor analysis, and the computation of the Multi-trait Genotype-Ideotype Distance Index.In the initial stage, each trait (rXij) was adjusted by: Where ߟ and ߮ are the new min and max values for the trait j. ߟ dan ߮ are the initial min and max values for the trait j, ߠ is the initial value of trait j genotype i.For the traits in which positive gain are desired ߟ = 100 dan ߮ = 0.The next step is factorial analysis using following equation: Where ‫ܨ‬ is matrix g x f with factorial score, ܼ is matrix g x p which consist of rescaled means, ‫ܣ‬ is matrix of ononical loading, R is correlation matrix of p x p, g, f, p are number of genotype, factor retained, and observed trait.The final stage involves the computation of the Euclidean distance between genotype and the ideotype scores, which is defined by the MGIDI index using the following equation: Where ߛ is is the score of the i th genotype in the j th factor, ߛ is the j th score of the ideal genotype.A low MGIDI index indicates the closeness of the genotype to the ideotype and thus indicates desired values for all the measured traits.The proportion of MGIDI index in i th genotype in the j th factor ߱ is used to describe the strength and weakness of each genotype using following equation: Where ‫ܦ‬ ଶ is a distance between i th genotype in the j th factor

Estimation of Genetic Parameter
Estimation of broad-sense heritability (h 2 ) was estimated according to [14] as the genotypicphenotypic ratio with formula: Selection gain and selection differential was calculated according to the formula as describe by [15] Xsi is the mean value of individuals selected for trait i; Xso is the initial population mean; SD is the differences between selected population and initial population mean; and h² is heritability of trait i.
The selection intensity used in this study was 15%.

RESULT AND DISCUSSION
The genotype selected by the MGIDI index were DT11, DT6, DT1 (Figure 1).Crossing number DT11 and DT6 are descendants of Deli dura with recombination of La Me x Rispa pisifera.DT11 and DT6 were produced from female parents which had been used as parents of commercial varieties, namely DxP PPKS 540.The advantage of DxP PPKS 540 was the very high percentage of mesocarp per fruit (88-90%).Based on the mean performance, the percentage of mesocarp per fruit DT11 and DT6 was still lower (83.8 and 87.8%) than the existing variety.However, these two progenies have IER yield potential of 28-30% (data not shown), 2-4% higher than the existing varieties.In addition, based on their vegetative characters, DT11 and DT6 had lower height increment (72-73 cm) and rachis length (5.1-5.2 m) compared to DxP PPKS 540 [16].This indicated that these two crosses demonstrated compact plant performance.One explanation for these differences is heterosis, where oil palm crosses are likely from combined dominance gene action [17].The La Me population is known for its smaller bunch and fruit size, high bunch production, and lower height increment [18] [19].The SP540 variety is recognized for its robust growth, early fruiting, slender shell, dense mesocarp, and advantageous quality of producing a high amount of oil [20].Recombination of La Me and SP540 origin is a form for stacking desirable traits.DT1 is derived from crossing Deli dura with Binga pisifera, known for good bunch oil content, oil yield, and ratio mesocarp to fruit [21].The same thing was also found in this study.Based on the mean performance DT1 has good bunch weight (12.97kg), ratio mesocarp to fruit (83.79%), oil to dry mesocarp (79,67%), and oil to bunch ratio (31.60%).According to [22] Binga pisifera transmitted lower oil to bunch ratio (22.7%) and ratio mesocarp to fruit (80.5%).It is assumed it was the female parent's influence.The female parent is a descendant of recombination between Tinjowan and Gunung Bayu, which have been used in producing the commercial variety, known by good ratio of oil to mesocarp and fresh fruit bunch.Based on the vegetative trait, the height increment of DT1 is 76.60cm/year.DT1 is slightly higher than DxP PPKS 540, but the height increment is still in the range of DxP Langkat, which means DT1 has the potential to have a compact palm trait.Apart from these genotypes, it is worth noting that DT15 exhibited a proximity to the cut point, indicating that it may be in close proximity to the predetermined criteria.In breeding programs, the integration of multiple traits can become difficult when desirable traits work against each other.The MGIDI index reduces 12 traits into four groups based on correlation and relationship between traits, as shown in Table 1.The first factor (FA1) includes ODM, OWM, HI, and DIAM, which are biologically sensible correlations as robust vegetative growth can aid in capturing more sunlight for photosynthesis.FA2 consists of BW and FB, which is fruit-related trait.FA3 consists of MF, OB, SF, KF, LAI, which are antagonistic to each other.In general selection gain in all traits is low.It is assumed to occur due to the use of intensively selected materials.Strength and weakness in the selected genotypes show low contribution of FA1 and FA3 in DT1, low contribution of FA1 in, and low contribution of FA1 and FA2 in DT1.Generally, the contribution of FA1 in all genotypes is very low indicating most genotype have vigorous vegetative traits and lower oil content.These contributions may be utilized to select potential parents for future breeding.In this instance, cultivar development targeting plants with high mesocarp to fruit and oil to bunch ratios may include genotypes DT15 and DT18, aside from the selected genotype.

CONCLUSION
MGIDI could help breeders to choose an ideal ideotype to develop new varieties.The genotype selected by the MGIDI index were DT11, DT6, DT1 which mean this genotype is the most likely to have compact palm and good production trait.Beside from the selected genotype DT15 was very close to the cut point, which suggests that this genotype is close to predetermined criteria.The view on strength and weakness indicated that the oil palm breeding for the next cycle should direct effort on reducing the height increment and increasing oil to dry mesocarp and oil to wet mesocarp ratio.

Figure 1 .
Figure 1.Genotype ranking in ascending order.The red circle showed the cut point according to the selection pressure

Table 1 .
Factors linked to correlated trait, heritability and predicted selection gain