Analysis of macro nutrient related growth responses using multivariate adaptive regression splines

Abstract

Strawberry micropropagation is generally based on Murashige and Skoog mineral salts, and many cultivars grow well on this medium. However, the diverse species found in germplasm collections often do not thrive, which indicates a need to optimize the mineral nutrients. In this study, Multivariate Adaptive Regression Splines (MARS), was employed to predict shoot quality, multiplication, and leaf color responses of three strawberry species in response to the major tissue culture nutrients by generating functional associations. MARS is a non-parametric approach that can be used to deal with continuous and categorical data without requiring the strict distributional assumptions of the basic linear models. The MARS algorithm is capable of capturing non-linear patterns between the input and target variables. NH4NO3, CaCl2·2H2O, MgSO4·7H2O, KNO3 and KH2PO4 were tested in a range of 0.5 × to 3 × MS medium, within a computer-generated optimal design that consisted of 32 treatment combinations. The plant responses were affected by all of the major salts tested and the genotype factor. Multivariate Adaptive Regression Splines captured the significant factors and their interactions to predict optimal major salts suitable for all three strawberry species: 3300 mg L−1 NH4NO3, 862.4 mg L−1 CaCl2, 1110 mg L−1 MgSO4, 3439 mg L−1 KNO3, and 329.8 mg L−1 KH2PO4. This study identified the major nutrient needs of the three strawberry species and provides an alternative statistical technique for tissue culture data analyses. The MARS statistical approach was used to predict macro nutrient related growth responses of three strawberry species. The objective of the study was to make a gentle introduction to the MARS algorithm and show its potential application to tissue culture research.