Figure 1

Shannon entropy calculation for various tissue-specific expression patterns. Synthetic expression patterns are shown. Original expression data represented by black circles are processed by equation 2 using Tukey biweights (dashed line). Processed data are represented by red circles. Specific expression observations detected to be outliers by [3] are highlighted. Numbers in black and red indicate Shannon entropy scores for the original and the processed data, respectively. Shannon entropy can range from (a) 0 to (b) 3.32 in this case (logarithm to the base 2 of 10). Expression patterns such as (a) and (c) are defined as tissue-specific genes in a narrow sense. Tissue-specific genes in a broad sense include various expression patterns such as (a, c, d) up-type, (e) down-type, and (f) mixed-type. By virtue of data processing, ROKU can detect tissue-specific genes in a broad sense. Meanwhile, ROKU gives relatively high scores (close to 3.32) for non-specific gene expression patterns such as (g) and (h).