Semi-supervised peak calling solution

Status: Active

The peak calling routine presents a significant challenge in working with large scale human epigenetic data: while ULI-Chip-seq protocol generally allows for robust peak calling, it is considerably more variable than the conventional Chip-seq approach. Accurately dissecting the situation when both background and signal can vary is generally a prohibitively complex task for the unbiased peak calling approaches.

We propose the novel semi-supervised approach to peak calling. Fast and effective semi-supervised peaks analyzer SPAN Peak Analyzer is a multipurpose peak caller capable of processing both conventional and ULI-Chip-seq tracks. In the semi-supervised approach, the user annotates a handful of locations as peaks, valleys, or peak shores, and then uses these annotations to train the model that is optimal for a given sample.

SPAN Peak Analyzer - semi-supervised peak analyzer

  • Supports semi-supervised peak calling with JBR Genome Browser
  • Works with both conventional and ULI-ChIP-seq data
  • Works with both narrow and wide modifications
  • Supports optional control track
  • Capable to process tracks with different signal-to-noise ratio
  • Supports replicates on model level
  • Experimental: differential peak calling

One of the major challenges of semi-supervised learning is the procedural complexity of the manual annotation of the data, which often leads to inaccuracies and mix-ups. We created a brand new visualization tool that would support classical genome browser capabilities and provide readily accessible integrated peak annotation and peak calling capabilities. JBR Genome Browser allows one to upload tracks of the Chip-seq data and perform on-the-fly annotation and peak calling for a given annotation, where results of the peak calling are immediately visualized as BED files in the same session. It can serve not only as an efficient semi-supervised peak calling engine but also as a next-generation genome browser. It provides enhanced capabilities of viewing large files, observing multiple locations simultaneously, and gathering track statistics.

JBR Genome Browser - fast and reliable genome browser

  • Supports semi-supervised peak calling with SPAN Peak Analyzer
  • Support for common track formats, e.g. WIG, BED, etc.
  • Capable to open saved IGV sessions
  • Support for combining tracks, i.e. arithmetic operations
  • Easy visualisation of multiple genomic regions
  • Custom statistics tracks for multiple regions
  • Frontend-agnostic, can be used as desktop application and web-service

Scheme

5ea1561a40553.png

Summary

SPAN Peak Analyzer and JBR Genome Browser can be used separately as general-purpose peak caller and genome browser, respectively. However, together, they can serve as a complete solution for peak calling. The semi-supervised peak calling approach is capable of improving peaks consistency in datasets with different signal-to-noise ratio, as well as obtaining the best peak calling results for individual samples.


Advantages of integrative semi-supervised peak calling solution:

  • Ability to process both narrow and wide histone modifications
  • Support for both conventional and Ultra Low Input Chip-seq experiments
  • Capable processing tracks with different signal-to-noise ratio
  • Call peaks and explore results in a single application
  • Integrated labeling functionality
  • Parameters optimization is super fast after the SPAN model is trained
  • Consistent peak calling among samples

A comprehensive step-by-step tutorial on using an integrated peak calling solution is available in How To section.

Participants