Fig. 2
From: Distributed gene expression modelling for exploring variability in epigenetic function
Hierarchical clustering of cell-lines by mRNA transcript abundance residuals (\(\varepsilon = Y - \hat {Y}\)). The three mesodermal derivatives GM12878, K562 and HUVEC cluster together, suggesting that residuals are partially non-random and instead convey meaningful biological information. Consistently, it is evident that the expression levels of many genes are poorly predicted across all eight cell-lines, presumably capturing divergence from histone modification-mediated regulation (explored in detail in our previous study [2])