Linking hyperspectral imaging data with gene expression profiles.

Healthy and B. graminis f.sp. hordei inoculated primary leaves were used for hyperspectral imaging and RNA extraction. After hyperspectral image acquisition and normalization, the Savitzky-Golay-filter was applied to reduce measurement noise. Reflectance values were manually extracted from regions of interest and used to identify relevant time points in the barley-Bgh interactions by applying the Relief algorithm. Spectral reflectance of identified relevant time points were used to calculate spectral differences between healthy and Bgh inoculated barley. In parallel, RNA was used for reverse transcriptase to synthesize cDNA and to analyze gene expressions of selected genes using real time PCR. Linear Discriminant Analysis (LDA) was applied on the hyperspectral data, gene expression profiles and a combination of both data sets to visualize the suitability of the features for the separation of different barley-Bgh interactions. To determine relevance as indicator for interpretable relations between hyperspectral signatures and gene expression profiles the Relief algorithm was applied (blue = high relevant wavelength band).