Abstract
AbstractThe large diversity of functional genomic assays allows for the characterization of non-coding and coding events at the tissue level or at a single-cell resolution. However, this diversity also leads to protocol differences, widely varying sequencing depths, substantial disparities in sample sizes, and number of features. In this work, we have specifically designed a suite of tools for exploring the non-coding genome, particularly for identifying consensus peaks in peak-called assays, as well as linking non-coding genomic regions to genes and performing Gene Set Enrichment Analyses. We demonstrate that a generic but flexible count modelling approach can be utilised to compare different conditions across a broad range of genomic assay such as ENCODE H3K4Me3 ChIP-seq, scRNA-seq and TCGA ATAC-seq. Our Python package, MUFFIN, offers a suite of tools to address common issues associated with high-dimensional genomic data, such as normalisation, count transformation, dimensionality reduction, differential expression, and clustering. Additionally, our tool integrates with the popular Scanpy ecosystem and is available on Conda and athttps://github.com/pdelangen/Muffin.
Publisher
Cold Spring Harbor Laboratory