Fig. 1

Heterogeneity of gene expression within and among disease sub-types. a Two-way hierarchical clustering using Ward’s method of standardized normal (z-scores) of transcript abundance of 11,614 genes (columns) in 202 individuals (rows). Six clusters identified to the right group individuals with similar profiles with respect to at least nine clusters of co-expressed genes. Letter beneath the heat map highlight BIT corresponding to genes enriched in reticulocytes (R), neutrophils (N), B cells (B), T cells (T), or for the interferon response (I). b Proportion of individuals of each disease sub-type represented in each of the six clusters of individual. For example, 45% of the UC samples are in the dark blue cluster, 30% in the red, 20% in the green, and 5% in the pale green, with none in the brown or light blue. c Principal variance component analysis shows the weighted average contribution of disease, sub-type within disease, or age-of-onset before 6 to the first five PC (67%) of the total gene expression variance, with the remainder residual variance unexplained, including individual differences