We have reported (MNL 59:77) on the variation observed among the arrays of polypeptides synthesized by developing maize embryos. Qualitative differences in the species of polypeptides synthesized and quantitative differences in the levels of synthesis were detected among the fluorograms obtained from embryos of different ages. Recent work has led to the adaptation of multivariate techniques for use in the analysis of patterns resulting from the two dimensional electrophoretic separation of polypeptides (Fewster and Walden, Computers in Biology and Medicine, in press).
The method involves the division of a stained gel or fluorogram into a number of discrete grid units each of which represents an axis along which variation may occur. By established criteria, we have selected 16 grid units for the study summarized below. The positions of the lines forming the grid are determined with respect to those of a number of widely separated, readily identified polypeptides such that the entire grid structure may be reproduced accurately on all gels included in the analysis. Each grid unit is assigned a numerical value which reflects the number and intensity of spots residing therein. A variety of different analytical procedures may be employed subsequently to examine the data.
Examples of the application of these methods to a study of variation in polypeptide synthesis during maize embryogenesis are presented in Tables 1 and 2. The data were obtained from the two dimensional arrays of polypeptides synthesized by embryos of Oh43 at 15, 20, 25, 30, 35, 40, 45 and 52 days after pollination. Table 1 shows the results of a Principal Components Analysis (PCA) where the first two components (PC1 and PC2) comprised 66.6% and 15.5% of the total variation respectively. The greatest contributions to the variation expressed by these components was evident in eigenvectors 3, 5, 7, 10 and 11 which represent the corresponding grid units. PC1 appears to contrast variation in grid unit 5 with that in grid units 3, 7, 10 and 11. Upon examination of the distribution of individual component scores along this component, a developmental trend emerges in which the density of grid unit 5 is observed to increase and the density of grid units 3, 7, 10 and 11 decrease as embryogenesis progresses between 15 and 55 days. PC2 expresses variation attributable to the combined effects of grid units 5 and 10. In this case a discrimination is made between embryos mid way in development from those which are older and younger, in that the combined relative densities of these two grid units are higher in the former.
Table 2 shows results obtained from a concentration analysis of these data where deviations from random expectation are measured for all grid units of an individual. This analysis partitions the variation into a number of lattices, the first of which is presented here, comprising 57.5% of the total.
The results obtained suggest a developmental trend similar to that which emerged through PCA. As embryogenesis proceeds from early through late development, the deviations associated with grid unit 5 change from negative to positive while the reverse is observed from grid units 3, 7, 10 and 11, indicating shifts from lower to higher and higher to lower densities respectively. Values obtained for these deviations in embryo mid-development are intermediate to those of early and late development.
Although some variation was lost in the process of grouping polypeptides into grid units, good correlation was observed between results obtained with this method and those of conventional forms of analysis. The multivariate techniques described should prove effective in the detection and representation of trends in any investigation in which the analysis of two dimensional separation of molecules is undertaken.
Tables 1 and 2.
J.G. Boothe and D.B. Walden
Return to the MNL 61 On-Line Index
Return to the Maize Newsletter Index
Return to the Maize Genome Database Page