371 Classification of the number of active starch synthase genes by NIR spectroscopy.

S. R. DELWICHE (1) and R. A. GRAYBOSCH (2). (1) USDA, ARS, Beltsville Agricultural Research Center, Beltsville, MD 20705; (2) USDA, ARS, Lincoln, NE 68583.

Newly proposed low amylose wheat varieties, currently under development, will have unique processing characteristics, and thus allow millers to blend defined levels of amylose in mixes requiring low- amylose flour. The amount of amylose synthesized during grain fill is dependent on the expression of three structural genes that encode isoforms of granule-bound starch synthase (GBSS). Lines possessing all three waxy loci as fully functional produce the highest proportion of amylose to amylopectin, while those with one or more null alleles produce successively smaller proportions. Conventional methods for determining the number of active starch synthase genes or isoforms, such as SDS-PAGE, are difficult and time consuming. The present study was undertaken to determine the feasibility of using NIR spectroscopy to classify wheat by the number of active GBSS genes. Nearly 200 lines of 1998-harvest wheat with 0–3 active genes were scanned (1100–2500 nm) in ground- and whole-grain form. Classification algorithms, such as SIMCA on PCA scores (Euclidean or Mahalanobis distance metrics), PCA of one population space, and PLS regression (0 = no active genes, …, 3 = all genes active), demonstrated that perfect separation of fully waxy lines was achievable, regardless of algorithm and sample form. SIMCA PCA with a Mahalanobis distance metric was most effective at separating intermediate cases (>80% accuracy). Misclassifications were most often assignments into neighboring gene classes (e.g., 1-gene line assigned to the 2-gene class). Models will become refined with incorporation of 1999-harvest samples.