Analysis and prediction of super-enhancers using sequence and chromatin signatures

Background: Super-enhancers are clusters of active enhancers densely occupied by the Mediators, transcription factors and chromatin regulators, control expression of cell identity and disease associated genes. Current studies demonstrated the possibility of multiple factors with important roles in super-enhancer formation; however, a systematic analysis to asses the relative contribution of chromatin and sequence features of super-enhancers and their constituents remain unclear. In addition, a predictive model that integrates various types of data to predict super-enhancers has not been established. Results: Here, we integrated diverse types of genomic and epigenomic datasets to identify key signatures of super-enhancers and their constituents and to investigate their relative contribution. Through computational modelling, we found that Cdk8, Cdk9 and Smad3 as new key features of super-enhancers along with many known. Comprehensive analysis of these features in embryonic stem cells and pro-B cells revealed their role in the super-enhancer formation and cellular identity. Further, we observed significant correlation and combinatorial predictive ability among many cofactors at the constituents of super-enhancers. By utilizing these features, we developed computational models which can accurately predict super-enhancers and their constituents. We validated these models using cross-validation and also independent datasets in four human cell-types. Conclusions: Our analysis of these features and prediction models can serve as a resource to further characterize and understand the formation of super-enhancers. Taken together, our results also suggest a possible cooperative and synergistic interactions of numerous factors at super-enhancers and their constituents. We have made available our analysis pipeline as an open-source tool with a command line interface at https://github.com/asntech/improse.