Click on image to view larger version.
Figure 3. Identification of putative maize mutants with altered spectrotypes based on class modeling. Principal component (PC) analysis of the near infrared reflectance (NIR) spectra of W22 wild-type leaves resulted in a class model based on three PCs (explaining 55, 29, and 7% of variance, respectively), that defined the wild-type. The PC scores of individuals from segregating families were calculated with this class model and evaluated using 
2 statistics based on squared Mahalanobis distances. The three graphs represent score plots based on principal components 1 (horizontal axis) and 2 (vertical axis) obtained from (A) a family without any outlier spectra; each black dot represents one individual family member; (B) a family containing several individual plants with spectrotypes that are not consistent with W22 spectra (indicated by black triangles, squares, and diamonds); and (C) the same family as in panel B, plus a related family (progeny of a sibling of the parent that gave rise to the family in panel B). Note how the second family contains individual plants with spectrotypes that are also inconsistent with W22 spectra, but similar to a cluster of outliers from the first family (black triangles). This provided evidence for the presence of a genetic mutation that alters the NIR spectrotype. The same scale was used in all three graphs. The axes intersect at the origin.
