FiberBrowser: a unified framework for single-molecule genomics

Single-molecule footprinting techniques are revolutionizing chromatin biology. These techniques allow for the identification of nucleosome positioning, regulatory-factor occupancy, and accessibility on individual DNA fibers, revealing a level of heterogeneity that bulk measurements fail to capture. A growing number of these footprinting methods utilize various enzyme systems and sequencing platforms.

To accommodate these diverse systems, researchers have developed FiberBrowser, a unified platform designed to transform such datasets into a common per-molecule representation of nucleosomes, accessible DNA, and subnucleosomal footprints. This representation enables interactive and scriptable analyses. The core of FiberBrowser is its inference engine, FiberHMM, which assesses each observation against the assay's informative sites and sequence-specific modification probabilities.

By distinguishing between missing information and protection, FiberHMM effectively separates these factors in four assays. FiberBrowser excels in clustering, sorting, and comparison of these calls across thousands of molecules in mere seconds. This enables researchers to link chromatin architecture to haplotype, CpG methylation, and non-reference structural variants within the same molecules.

Importantly, the platform's representation is chemistry-independent, allowing for the transfer of models between assays. A notable feature is FiberBrowser's ability to classify cells from Fiber-seq footprints, correctly identifying 12 scDAF-seq cells as lymphoblastoid without the need for additional training. In contrast, models trained on raw modification patterns struggle to transfer to new assays, highlighting FiberBrowser's versatility and expandability in the field of single-molecule chromatin footprinting.