Submitted to: Cereal Conference Royal Australian Chemical Institute Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 8/16/1998
Publication Date: N/A
Citation: N/A Interpretive Summary: Wheat-rye chromosomal translocations, designated as 1AL.1RS or 1BL.1RS, particularly those involving the short arm of rye chromosome 1R, have been used during the past 20 years to impart wheat hardiness and resistance to pathogens and insects, broaden adaptation, and increase yield. However, the presence of the 1AL.1RS or 1BL.1RS rye translocations in wheat has been shown to result in inferior dough handling and baking characteristics. Although numerous analytical techniques (e.g., HPLC, monoclonal antibody tests, high performance capillary electrophoresis) have been developed for detecting these translocations, the complexity of the analytical procedures restricts their use to research and analytical laboratories. In previous research, we reported on the development of near-infrared based regression models that are capable of correctly classifying wheat by presence or absence of the 1RS translocation to within 98% accuracy. The current research builds upon these results by attempting to understand the spectroscopic differences that naturally arise between samples with 1RS and those without. Two methods of cluster analysis are used: principal components analysis and artificial neural networks Kohonen mapping. Results indicate that clustering into 1RS and non-1RS groups is not evident when the spectral data is compressed to a mathematical dimension of two or three. However, the cluster analysis schemes are effective in displaying dominant spectral features such as protein content and year of growth. This work is intended to benefit wheat breeders, who seek a rapid and robust means of identifying baking quality characteristics in wheat.
Technical Abstract: Detection of wheats that possess the 1AL.1RS or 1BL.1RS wheat-rye translocation is currently possible through means of HPLC, electrophoretic methods, and immunoassay. Adaptation of these methods to commercial grading, inspection, and processing facilities is not likely because of their complexity, high cost, and long analysis times. As an alternate method, near-infrared (NIR) spectroscopy has been proposed for 1RS identification because of its rapidity, ease of use, and widespread acceptance by the cereals industry. Uncertain at the time of development were the biochemical or physical properties, that are spectrally sensed, which account for the ability of such classification. The current study has examined the existence of spectrally derived patterns in a diverse set of wheat [which consists of 30 cultivars (8 of which possess the 1RS translocation), 8-9 geographical locations, and 2 growth years] without prior knowledge of such genetic and cultural differences. Two methods were used: principal components analysis and the self organizing artificial neural network (ANN) of the Kohonen type. Results indicate that, despite reasonably high to high classification accuracies for corresponding classification algorithms [principal components regression (87%), feedforward back propagation ANN (97%)], the most discernible feature, among examined choices of the presence/absence of 1RS, geographical origin, cultivar and crop year, was overall protein content. However, because of the manner in which 1RS and non-1RS samples were selected for self-organization and classification, it is not believed that this feature is responsible for the success of 1RS classification by NIR spectroscopy.