Skip to main content
Premium Trial:

Request an Annual Quote

This Week in Genome Biology: May 16, 2018

A Peking University-led team shares details on its antibody-independent chromatin interaction mapping method, known as open chromatin enrichment and network Hi-C (OCEAN-C). The approach combines in situ Hi-C assay and FAIRE-seq methods, the researchers explain, using active cis-regulatory element patterns to flag open chromatin interactions. With it, they tracked down some 10,000 hubs of open chromatin interactions (HOCI) in human multiple myeloma or lymphoblastoid cell lines, which they went on to validate and characterize with new and available chromatin, genome structure, and transcriptome data. "We demonstrate that HOCIs preferentially form open chromatin interactions, including promoter-enhancer, promoter-promoter, and enhancer-enhancer interactions," the authors note, "which distinguish HOCIs form other open chromatin regions." 

Researchers from Tel Aviv University describe an inference method for dialing down false discovery rates when predicting enhancer-promoter interactions. The FOCS approach — named for the FDR-corrected OLS with Cross-validation and shrinkage strategy behind it — considers enhancer-promoter links "based on correlated activity patterns across many samples from heterogeneous sources," the team explains. By applying FOCS to thousands of samples profiled for ENCODE and other large projects in the past, the investigators tracked down roughly 300,000 enhancer-promoter interactions involving around 16,000 genes. Using raw GRO-seq data in the Gene Expression Omnibus database, meanwhile, they used it to put together an enhancer RNA and gene expression compendium for 245 samples representing nearly two dozen human cell lines.

A research duo from Dana-Farber and Harvard introduces GiniClust2, a computational approach for identifying common and rare cell types from single-cell data. The GiniClust2 pipeline brings together complementary clustering methods, the authors note: the Gini index for detecting rare cell types based on differentially expressed genes and a Fano factor better suited to more common cell types. "Instead of averaging results from individual clustering methods, as is traditionally done, GiniClust2 selectively weighs the outcomes of each model to maximize the methods' respective strengths," they write. After testing GiniClust2 on simulated single-cell RNA sequence data, the team applied it to single-cell data for 68,000 peripheral blood mononuclear cells previously profiled by 10X Genomics, uncovering nine common and two rare cell types. They also showed that it could detect rare cell types in differentiating mouse embryonic stem cells.