Association Between Agronomic and Biomass Component Traits Using Pathway Analysis and Regression Tree Techniques in Biomass Sorghum

Abstract

The aim of this study was to estimate the correlations among traits in biomass sorghum, estimate the direct and indirect effects, and check for a cause/effect relationship of the traits on the theoretical calorific value (TCV) through pathway analysis. An additional aim is to provide a complementary technique for analysis of the traits and orientation of breeding programs of biomass sorghum through data mining. A total of 25 hybrids of biomass sorghum were evaluated in a triple square lattice design. The following traits were measured: flowering (FLOW); plant height (PH); fresh biomass yield of the stalk (FBYs), of leaves, and of panicles; dry matter of the stalk and of the leaves (DMl); neutral detergent fiber of the stalk and of the leaves; acid detergent fiber of the stalk and of the leaves; lignin of the stalk and of the leaves; hemicellulose of the stalk and of the leaves; cellulose concentration of the stalk and of the leaves; and the TCV. The data were placed under exploratory analysis, the mixed model approach, pathway analysis based on genotypic correlations, and the regression tree technique. The traits FLOW, PH, FBYs, DMl, and LIG were considered the main determinants of the changes in the TCV and, therefore, with potential for use in indirect selection. The technique of data mining allowed better visualization of the relationships among the traits, assisting in some explanations regarding the effect of the traits on the TCV.