DNase-seq Data Standards and Processing Pipeline

Pipeline Overview

The ENCODE 4 DNase-seq pipeline was developed in collaboration with the Stamatoyannopoulos lab as a part of the ENCODE Uniform Processing Pipelines series. The pipeline for processing DNase I sequencing data takes in Illumina sequencing reads from DNase I experiments, assess the data, and returns genomic regions with statistically significant enrichments, or 'hotspots', of cleavage activity from the original DNase I experiment. The full DNase-seq pipeline code is available on Github.

DNase-seq pipeline

View the current pipeline instance.

Inputs:

File format	Information contained in file	File description	Notes
fastq	reads	G-zipped reads, paired-ended or single-ended.	Multiple FASTQs from a single biological replicate or library are concatenated before mapping
fasta	genome indices, genome reference	Indices are dependent on the assembly being used for mapping
txt	chromosomes reference	List of expected chromosomes
bed, starch	hotspots reference	Reference files used for hotspots and peaks calling	Hotspots reference must be selected guided by the input FASTQ read length
txt	bias model	Reference file used for footprints modeling

Outputs:

File format	Information contained in file	File description	Notes
bam	unfiltered alignments	FASTQs are concatenated, trimmed for adapters and length, and aligned with BWA
bam	alignments	unfiltered alignments are filtered and deduplicated
bigWig	read-depth normalized signal	Density starch from Hotspot2 is normalized and converted to a bigWig
bed and bigBed narrowPeak	peaks	DNase I hypersensitivity sites identified using Hotspot2	Both 5% and 0.1% FDR peak calls are included in the outputs. FDR stands for False Discovery Rate.
bed and bigBed bed3+	FDR cut rate	FDR estimate per site for being called hypersensitive to cleavage by DNase I
bed and bigBed bed3+	footprints	A footprint model is fit and statistical deviations from expected cleavage rates within hotspot regions are called and thresholded	1% FDR footprints. A detailed explanation of the footprint modeling is available in the footprint-tools documentation.
starch	peaks	DNase I hypersensitivity sites identified using Hotspot2	0.1% FDR peaks

References

Genomic References

View the reference files used in this pipeline

Links and Publications

Find data generated by the ENCODE 4 DNase-seq pipeline

Back to the top

Current Standards

Experimental guidelines for DNase experiments can be found here.

• Experiments should have two or more biological replicates, isogenic or anisogenic. Assays performed using EN-TEx samples or other rare types may be exempted due to limited availability of experimental material.
• A SPOT (Signal Portion of Tags) score of 0.4 or higher is considered a product of high quality data.
  • A SPOT score of 0.25 is considered minimally acceptable for rare and hard to find primary tissues.  In very rare cases of limited sample availability, lower scoring data may be used with appropriate caution. 
  • Any sample with a SPOT score <0.3 should be targeted for replacement with a higher quality sample, whenever possible.
  • SPOT scores should be calculated on de-duplicated data or on data for which the duplicate rates are <5%. 
• DNase-seq requires a minimum of 20 million uniquely mapping reads to generate a reliable SPOT score, and 100 million uniquely mapping reads to generate reliable DNase footprints.
  • For a standard conventional DNase-seq profile, 50 million uniquely mapping reads are recommended.
  • For deep, footprinting depth DNase-seq, a depth of 150-200 million uniquely mapping paired-end reads are recommended.
• Acceptable mappability rates and mitochondrial content are listed below:

SPOT Score*	Mitochondrial Fraction	Mappability (without mitochondrial fraction)	Resulting Data Status
≥ 0.4	< 10%	> 75%	Good
0.3 - 0.4	10 - 15%	65 - 75%	Acceptable
< 0.25	> 15%	< 65%	Poor
* SPOT scores of 0.25 - 0.3 are minimally acceptable in certain circumstances (see above explanation)

• Replicate concordance: the gene level quantification should have a Pearson correlation of >0.9 between isogenic replicates and >0.85 between anisogenic replicates.
• The experiment must pass routine metadata audits in order to be released.

Uniform Processing Pipeline Restrictions

• The read length should be a minimum of 36 base pairs.
• Read trimming of adapter sequences is recommended.
  • Failure to trim sequences of fragments sized shorter than the sequencing cycle number can lead to filtering small fragment signal and create bias in the resulting alignments.
  • There is no post-trimming read length minimum, but effective fragment mapping drops significantly at kmer-lengths of less than 22 base pairs.
• Adapter sequences used in library creation should be documented and available to the pipeline.
• Barcodes/UMI coding should be indicated in the metadata and available for accurate application of duplication filtering methods by the pipeline.
• Sequencing may be paired- or single-end, as long as sequencing type is specified and read pairs are indicated. Paired-end sequencing is preferred. 
• The sequencing platform used must be indicated. 
• Alignment files are mapped to either the GRCh38 or mm10 sequences.

Back to the top