Cucurbit Genetics Cooperative Report 16:27-29 (article 9) 1993
Application of factor analysis to cucumber breeding
Meng Zhang and Hongwen Cui
Department of Horticulture, Northwestern Agricultural University, Yangling Shanxi, 712100, P.R. China
Cucumber breeding objectives suggested by the "Chinese Cucumber Breeding Cooperation Team" include the development of an early maturing, high yielding cucumber which is resistant to anthracnose, fusarium wilt and downy mildew disease, and which posesses good commodity characters for Chinese markets. These objectives require the incorporation of many quantitavely inherited traits.
It is necessary to develop a comprehensive selective method for reaching these objectives. Studies which identify genetic relationships among traits (correlations) is a prerequisite for any comprehensive selection method. there are certain limitations inherent in traditional correlation analysis. Factor analysis, however, cam be used to overcome some problems which are inherent in correlation analysis. The aim of this study was to investigate genetic relationships among several traits in cucumber using factor analysis.
An experiment was conducted at the Horticulture Station of the Northwestern Agricultural University. Twenty-four varieties and inbreds were evaluated in a randomized block design with 3 replications. Ten plants of each cultigen were randomly chosen to evaluate 26 quantative traits during the growth period. Traits included: 1) the node position of the first pistillate flower (x_{1} ); 2) the days from sowing to the first pistillate flowering plant in the population (x_{2}); 3) the days from sowing to first male flowering plant in the population (x_{4}); 5) the days from sowing to staminate flowering in 50% of the plants (x_{5}); 6) leaf area per plant in the early stage (x_{6}); 7) fruit length (x_{7}); 8) fruit diameter (x_{0}); 9) leaf number in the early stage (x_{8}); 10) pistillate flower density (main vine) in the early stage (x_{9}); 11) number of pistillage flowers (main vine) in the early stage (x_{10}); 12) number of staminate flowers (main vine) in the early stage (x_{11}); 13); number of harvested fruit per plant in the early stage ((x_{12}); 14) average fruit weight in the early stage (x_{13}); 15) early yield per plant (x_{14}); 16) total number of branches (x_{15}); 17) average length between two close nodes (x_{16}); 18) total number of leaves (x_{17}); 19) the highest fruit setting node (x_{18}); 20) number of harvested fruits per plant (x_{19}); 21) total fruit weight per fruit (x_{20}); 22) total yield per plant (x_{21}); 23) downy mildew resistance in the early stage (x _{22}); 24) downy mildew resistance in the late stage (x_{23}); 25) fusarium wilt disease incidence (percentage; x_{24}) and; 260 anthracnose disease incidence severity (x_{25}). Resistance to anthracnose was identified by an in vitro leaf method.
Original data of x_{9}, x_{23} and x_{24} underwent anti-sine transformation and the variance was analyzed in a randomized block design. Genotype values (g_{ij}) of traits were estimated according to Chui-Yu Liu (1981) and then genotype correlation coefficients were calculated on the basis of genotypic value.
The mainfactor solution (factor analysis) was calculated using the genotype correlation mateix. The orthogonal factor load matrix was calculated by orthogonal rotation and transformation of the initial maximum variance using the BLQMIN oblique rotation method. The oblique factor load matrix and oblique factor correlation matrix were also analyzed.
Variance analysis detected significant differences among all observed traits, except for fruit diameter (x_{0}). Results show that the difference between model parameters was caused by genetic factors. Therefore, the genetic analysis could be evaluated in 25 x 25 matrix (25 traits).
Most of the trait loads centralized on a factor (F_{1-5}, Table 1). A biological explanation of all factors can be given: in factor F_{1}, x_{6}, x_{8,}, x_{7}, x_{13} , x_{14}, x_{16}, x_{19}, x_{21}, x_{23}, x_{24} and x_{25} traits occupied the higher load. These traits centralized all yields and component factors. except the x_{12} trait. Therefore, F_{1}factor is named the "Yield factor". Yhtrr kinds of disease (anthracnose, fusarium wilt and downy mildew) had a higher load, which indicated that there was close association between plant resistance disease level and yield. All of the susceptible parameters were negatively loaded, which indicated that they are negative factors for yield.
Higher traits loads in factor F_{2} included x_{1} , x_{3}, x_{9}, x_{10}, x_{11}, x_{12} , and x_{14} which were associated with traits of earliness, and F_{2} was named the "early - mature factor". The node of the first female flower,pistillate flower density (main vine) in the early stage, and number of pistillate flowers (main vine) in early stage had very high loads. Results indicate that these traits had the greatest effect on earliness.
Higher trait loads infactor F_{3} included x_{6}, x_{8}, x_{17} and x_{18} which reflected plant nutrient level and plant growth vigor. Therefore, this factor was named the "Yield physiology factor". The load of total average fruit weight was high and average fruit weight at early stage also had a positive load. These data indicate physiological factors had a direct positive effect on fruit weight.
The load of x_{2}, x_{3}, x_{4}, and x_{5} was higher than other traits in factor F_{4} and was named the "flowering season factor". In addition, x_{6}, x_{8}, x_{7} and x_{12} traits also had a certain load. The number of harvested fruit was only positive value. Results show that components of this factor could result in delayed growth and development, while influencing yield by increasing fruit weight and reducing the number of fruit.
High trait loads were observed for x_{4}, x_{5}, x_{11}, x_{15}, x_{22} and x_{23} in factor F_{5}. This factor was named the "male flower development factor". The time of male flowering and the number of male flowers (increase) during the early partof the season was closely associated with downy mildew susceptibility.
Table 1. Factor loading matrix afgter orthogonal rotation transformation (genotype).
Traits |
F_{1} |
F_{2} |
F_{3 } |
F_{4} |
F_{5} |
X_{1} |
0.0909 |
-0.7137 |
0.3977 |
0.2239 |
0.1267 |
X_{2} |
0.3323 |
-0.3775 |
-0.0021 |
0.7427 |
0.2136 |
X_{3} |
0.1331 |
-0.5830 |
0.1883 |
0.5687 |
0.1845 |
X_{4} |
0.2033 |
0.0274 |
-0.0219 |
0.8166 |
-0.3887 |
X_{5} |
0.3820 |
-0.0692 |
-0.1084 |
0.7453 |
-0.3660 |
X_{6} |
0.7670 |
-0.0303 |
0.5123 |
0.3280 |
-0.1242 |
X_{7} |
0.8639 |
-0.0504 |
0.1894 |
0.3222 |
0.1484 |
X_{8} |
0.5668 |
-0.0169 |
0.6095 |
0.3367 |
-0.1954 |
X_{9} |
0.0943 |
0.9708 |
-0.0675 |
0.0824 |
0.1457 |
X_{10} |
0.1077 |
0.8622 |
0.0986 |
-0.0330 |
0.1400 |
X_{11} |
0.0516 |
-0.4363 |
0.3967 |
-0.0462 |
0.5504 |
X_{12} |
0.3366 |
0.7722 |
-0.1883 |
-0.3558 |
0.0500 |
X_{13} |
0.8306 |
-0.0427 |
0.2732 |
0.3145 |
0.1383 |
X_{14} |
0.8109 |
0.5093 |
-0.0292 |
0.0283 |
0.0666 |
X_{15} |
0.0012 |
-0.3291 |
0.1474 |
0.1302 |
-0.695 |
X_{16} |
0.7747 |
-0.1202 |
0.2234 |
0.0804 |
0.0307 |
X_{17} |
0.2910 |
-0.2044 |
0.8847 |
-0.136 |
0.0929 |
X_{18} |
0.4830 |
-0.1631 |
0.8038 |
-0.0436 |
0.1051 |
X_{19} |
0.8934 |
0.2664 |
0.1526 |
0.0530 |
-0.2160 |
X_{20} |
0.8191 |
-0.0834 |
0.4010 |
0.1978 |
-0.1020 |
X_{21} |
0.9173 |
0.0603 |
0.1689 |
0.1056 |
-0.2768 |
X_{22} |
-0.4571 |
0.2414 |
0.3065 |
-0.2842 |
0.5540 |
X_{23} |
-0.7594 |
-0.0988 |
0.1045 |
-0.1685 |
0.4936 |
X_{24} |
-0.8134 |
-0.1482 |
-0.2374 |
-0.2374 |
-0.1266 |
X_{25} |
-0.5227 |
-0.1296 |
0.1039 |
0.2561 |
0.3375 |
Factor analysis may be useful for describing and clarifying interrelationships among traits. As such this technique may be an important guide for selection during cucumber breeding.
Literature Cited
- Liu Chio-Yu. 1981. The genetic correlation and its addibility of crop quantitative characters. Anhui Journal of Agricultural Sciences. Special Edit for Quantity Genetics, 84-88.