Skip to content

Span Tuning

Jump to bottom
Roman Chernyatchik edited this page May 17, 2019 · 5 revisions

Span Tuning

This section shows JBR features required for semi-supervised peak calling in JBR.

Semi-supervised peak calling allows to get more consistent peak calling for noisy ChIP-Seq data and assumes 3 steps:

  • SPAN model fitting
  • Confident peaks annotation
  • Peak calling parameters tuning

For more information please refer to:

This page highlights only several JBR features required for the semi-supervised peak calling. You could learn how to do the whole pipeline at How-To On Semi-Supervised Peak-Calling.

Turn On Peaks Annotations Mode

Turn on peak annotations mode using Main Menu | Annotate | Peaks Annotation Mode or using toggle button.

alt text

Select Region

  • Hold Shift + Alt (MacOS: Shift + Option ⌥) keyboard key + Hold Left Mouse Button + Move Mouse to select the region (on Linux Shift + Alt could be a system shortcut and thus does't work here, as alternative you could use Shift + Control instead)

alt text

E.g. annotate as "Peak Start"

alt text

Annotate Selected Region

Select region then click on one of four label buttons or press s/e/n/p keyboard keys.

alt text

E.g. let's add Peak Start

alt text

Remove / Change Annotations

Annotations track provides context menu (accessible by Mouse Right Click) which allows to:

  • Select / Delete annotation under cursor
  • Change type using Peaks, No Peaks, Peak Start, Peak End context menu actions:
  • Show list of annotated chromosomes
  • Clear all annotations

alt text

Export & Import Annotations

Export / Import annotations from BED file using context menu Import from *.bed / Export to *.bed or using Import / Export buttons.

alt text

Load SPAN Model

Load SPAN model tracks using Main Menu | File | Load SPAN Model....

alt text

In order to visualize SPAN model track, JBR performs peak calling using default SPAN options parameters (FDR 1.0e-6, GAP 5). Default parameters could not match your data and you need to run tuning procedure to get the best result.

alt text

Tune SPAN Model

Select SPAN tracks and use Tune SPAN models context menu action to tune SPAN models on annotated regions. Also, you could use Main Menu | Annotate | Tune SPAN Models... action to tune all SPAN models loaded in JBR browser.

alt text

JBR shows progress while tuning tracks:

alt text

Finally, you will get tuned tracks:

alt text

Export Peaks from JBR

After tuning please export peaks to BED file using Export SPAN model peaks... action in SPAN track context menu or using Main Menu | Annotate | Export SPAN Peaks.... If one SPAN track is selected JBR suggests to specify an output file name. If several SPAN tracks are selected JBR asks to specify an output folder.

About Track

About SPAN track action includes additional information about mismatched annotations number for the tuned track

alt text

E.g.

============ Track (1 / 1): 'GSM3074494_BB_GVO_H3K36me3_mm9' ============
Track source: /Users/romeo/work/workshops/span_workshop2/workshop_files/GSE112622/span-models/GSM3074494_BB_GVO_H3K36me3_mm9.span
Source size: 50,6 mb
Peaks Statistics:
  Peaks number: 16 647
  Peaks summary length : 650 329 800 bp
  Genome (mm9 [all chromosomes]) coverage: 0.24%
  Lengths:
    Min: 200 bp
    Mean: 39 065 bp
    Max: 972 400 bp
    5%: 200 bp
    50%: 16 800 bp
    95%: 150 400 bp

  Track error score:
    Total error: 14% (6/42)
    No peaks error: 8% (1/12)
    Peaks error: 0% (0/10)
    Peak start error: 30% (3/10)
    Peak end error: 20% (2/10)
Clone this wiki locally