Abstract:
|
Hi-C is a powerful tool to investigate the 3D organization of the genome. A fundamental task in the Hi-C data analysis is to identify the interaction loci that form the base of chromatin loops. However, the interactions identified by current loop calling algorithms have low reproducibility across biological replicates, posing concerns on the reliability of identifications. One particular challenge in assessing the reproducibility of identified interaction loci across replicates is that the reproducibility of the identification is related to the proximity between the loci.
In this project, we propose a statistical method, called irreproducible discovery rate regression, for assessing the reproducibility of the interaction loci identified by loop calling algorithms and identifying reproducible interactions. Built upon the basic framework of irreproducible discovery rate (IDR) method, this method takes the aforementioned distance dependence effect into account in the reproducibility assessment. We demonstrate that the proposed method is able to identify loops with higher biological relevance.
|