Figure 3.

Features Contributing to Conservation of Promoter and Enhancer Activity Identified in Human Liver

(A) For all human proximal promoters active in liver, the depth of conservation was correlated with experimental features (reproducibility, peak intensity, peak length, distance to nearest transcription start site) as well as underlying genomic features (GC content, sequence constraint, TF binding sites). Each feature in isolation explained a significant fraction of the variance in conservation of promoter activity (e.g., peak length explained 10%). The fraction explained by the features in combination, when added left to right using multiple regression analysis, are plotted as a line above, in sum totaling 36%. The increases in explained variance with the addition of each feature are attenuated due to strong inter-correlation of features, quantified in the bottom panel as R² values between features (Experimental Procedures).

(B) The same analysis was performed for human liver enhancers, where experimental and genomic features together explained a more modest fraction (23%) of the conservation of enhancer activity in other species.