Comparison between predictable and unpredictable genes.

Compared features are the number of neighbors selected by the model (i.e., k), the absolute expression level of the gene, the coefficient of variation of the gene (i.e., standard variation normalized by mean), the number of SNPs per gene and the sum of entropy of SNPs per gene. Shown are the average values for each set of genes (850 predictable genes vs. 14,587 unpredictable genes in the top rows, and 412 K-nearest-neighbor-integrating-genomic-features (KNN-IGF) predictable genes vs. 90 simple-KNN predictable genes in the bottom rows), and the p-value calculated by Student's t-test (NS is non-significant).