New Method for Sugarcane (Saccharum spp.) Variety Resources Evaluation by Projection Pursuit Clustering Model

In the breeding of new sugarcane varieties, the survey data do not always conform with a normal or linear distribution. To apply non-normal or non-linear data to evaluate new material requires a suitable evaluation model or method. The projection pursuit clustering (PPC) model is a statistical method that does not require making normal assumptions or other model assumptions on sample data, and is suitable to analyze high-dimensional, non-linear, and non-normal data. However, this model has been applied infrequently to crop variety evaluation. In this study, 103 varieties that have been bred over the last 70 years in China were planted, and their main industrial and agronomic traits were collected. Through the exploratory analysis of the data structure characteristics, the PPC model was used to evaluate these sugarcane varieties. The model provided good projection directions of agronomic and industrial traits, with accurate projection values. PPC models could evaluate sugarcane resources well, and the results were objective and reliable. Thus, the PPC model could be used as a new method for crop variety evaluation. At the same time, 51 excellent industrial and agronomic variety resources were screened for application in breeding.

Automated summary

In breeding new sugarcane varieties, the traditional statistical methods may not be suitable due to the non-normal or non-linear distribution of the survey data. The projection pursuit clustering (PPC) model provides a promising approach as it can analyze high-dimensional, non-linear, and non-normal data without making assumptions. In this study, the PPC model was successfully used to evaluate sugarcane varieties and identify excellent resources for further breeding.