Exploring cis-regulatory models of the genome to predict epigenetic state and variation

[摘要] We introduce a new model called CCM+ that rectifies the inability of previous sequence-based methods to generalize across cell-types. We permit generalization by introducing arbitrary base-pair resolution covariates. We show that the addition of base-pair resolution chromatin accessibility covariate greatly aids in the prediction of cis-regulatory marks. Additionally, we show that by using cell-type specific covariates, CCM+ can generalize across cell-types. Finally, we show CCM+ can be used for downstream analysis that matches state-of-the-art methods when chromatin accessibility is used as a covariate.

[发布机构] Massachusetts Institute of Technology