Sequencing for epigenetics

https://doi.org/10.1016/B978-0-12-812215-0.00004-2

NGS techniques diversity

Types of binding events in the cell

Proteins and nucleic acids are the most important components of the cell. Their interactions are necessary for functioning and regulation. Types of interactions:

• DNA-protein interactions, examples:
  • replication/reparation/recombination
  • transcription
  • chromatin modification and folding
• RNA-protein:
  • RNA metabolism regulation
  • processing
  • translation
  • degradation: microRNA, NMD
• RNA-DNA interactions:
  • expression regulation
  • chromatin structure maintenance

Wasserman & Sandelin, 2004

Types of molecular interactions

• Direct:
  • protein-protein
  • RNA-protein
  • DNA-protein
  • RNA-DNA (R-loops)
• Indirect:
  • Protein-mediated
  • RNA-mediated
• Colocalization:
  • Compartments
  • Hubs

RNA-DNA interactions

Engreitz et al. 2016

ChIP-Seq

Binding of protein is typically specific to DNA sequence:

Binding pattern

DNA-protein binding

List of DNAs
with binding events

ChIP-Seq

Chromatin-immunoprecipitation followed by sequencing:

ChIP-Seq: input experiment

Unwanted factors contributing to sequencing & processing output:

• DNA accessibility
• DNA amplification efficiency
• Mappability

"Input" experiment is a ChIP-Seq control without precipitation step. It is used for normalisation of ChIP-Seq.

Park, Nat Reviews Genetics, 2009

ChIP-Seq: peak calling

In order to predict binging events we need to call peaks in ChIP-Seq data:

Mahony & Pugh 2015

ChIP-Seq: binding pattern

• As an output from ChIP-Seq we have a list of peaks.
• In theory, each peak should have a binding event.
• Imagine you have the peaks with the following sequences:
  
  
  
  
  
  
  
  
  
   
• What is the binding pattern?

ChIP-Seq: binding pattern

• Seems to be easy to find the binding pettern:

Question: Can you propose a computational approach to find it?

ChIP-Seq: motifs search

• Let's introduce some variety, which can arise as:
  • Suboptimal binding of the protein​
  • Genome variability
  • Errors during the sequencing

ChIP-Seq: motifs search

• With possible substitutions, finding the correct pattern is not easy anymore:

We need some advanced algorithms to find binding patterns.

Binding pattern representation

• Binding motif is a DNA sequence pattern that has biological significance (e.g. it is a target for protein binding).
• Consider following pattern of sequences:

TATAAT
TAAAAT
TAATAT
TGTAAT
TATACT

Consensus is the sequence of the most frequent letters:

T[AG][AT][AT][AC]T

With the consensus sequence, we lose the information:

• about suboptimal binding,
• of background nucleotides frequencies.

Binding motif representation

Other representations: 

Frequency matrix

Probability matrix

Matrix normalized by background

Position Weight Matrix (PWM)

Position Weight Matrix (PWM)

Binding motifs search 

Approaches:

•  Scanning of the motifs from the database
  (e.g. JASPAR http://jaspar.genereg.net/ )

Binding motifs search 

Approaches:

•  Scanning of the motifs from the database
  (e.g. JASPAR http://jaspar.genereg.net/ )
• Motifs discovery (de novo search)

http://meme-suite.org/tools/meme

ChIP-Seq: motifs search tools

• Web server tools:
  • http://rsat.eu/
  • http://meme-suite.org/
  • ...
• Command line interface tools (CLI):
  • MEME
  • HOMER
  • Autosome
  • ...
  • ​GimmeMotifs https://gimmemotifs.readthedocs.io/en/master/
    • Simple CLI
    • Can run multiple other tools
    • Databases included

DNase-Seq: experimental protocol

• DNase I hypersensitive sites sequencing

ATAC-Seq: experimental protocol

• Assay for Transposase-Accessible Chromatin with high-throughput sequencing
• Nucleosome-free regions detection method
• Tagmentation: Tn5 transposase cuts DNA and ligates adaptors to open DNA

Buenrostro et al., 2013

"Block-schemes" of Seqs:

Meyer & Liu Nat. Rev. Gen. 2014

Data processing of Seqs:

Meyer & Liu Nat. Rev. Gen. 2014

Examples of tracks:

Meyer & Liu Nat. Rev. Gen. 2014

In theory: 

ATAC-Seq

Hsu et al. 2018

paired-end sequencing and mapping

ATAC-Seq: insertion size distribution

Buenrostro et al., 2013

• Distance between forward and reverse mapping positions:

ATAC-Seq: insertion size distribution

Buenrostro et al., 2013

• Distance between forward and reverse mapping positions:

ATAC-Seq: insertion size distribution

Buenrostro et al., 2013

• Distance between forward and reverse mapping positions:

ATAC-Seq around gene starts

Buenrostro et al., 2013

• Average plot around
  Transcription Start Sites (TSS): 

ATAC-Seq around TF binding sites

 Albanus et al., 2019

We can plot V-plot around positions of factors binding sites:

ATAC-Seq around TF binding sites

 Albanus et al., 2019

ATAC-Seq around nucleosomes

Schep et al., 2015

We can also plot V-plot around positions of nucleosomes:

Chromatin openness (accessibility)

Hsu et al. 2018, Slattery et al. 2014

We can also do motifs search in open chromatin regions

Binding events and DNA motifs

Specific and non-specific binding:

Slattery et al. 2014

High-resolution antibody-targeted chromatin profiling

 Cleavage Under Targets and Release Using Nuclease (CUT&Run)

Cleavage Under Targets & Tagmentation (CUT&Tag)

High-resolution antibody-targeted chromatin profiling

 Cleavage Under Targets and Release Using Nuclease (CUT&Run)

Cleavage Under Targets & Tagmentation (CUT&Tag)

Genomics tracks association

DNase-Seq coverage for ∼350-kb region:

Clustering of samples

Let's check the similarity of pooled tissue samples with DNase-seq from known tissues. For that, let's plot bi-clustered heatmap of Spearman correlation coefficients:

Result of clustering,
the tree reflecting the similarity between samples

Clustering of samples

Finally, let's do dimensionality reduction, where each dot is a single sample.
Here is an example with t-SNE, although PCA (Principal Component Analysis) and MDS (Multidimensional scaling) are other popular techniques:

NGS data typical workflow example

Let's review the NGS data processing workflow:

1. Quality control: fastqc on initial FASTQ
2. Index build: bowtie2-build on initial chromosomes files
3. Read mapping: FASTQ → SAM/BAM
4. Filtering of mapped reads: samtools on SAM from p.3
5. Conversion to BED: bedtools on SAM from p.4

FASTQC report example

• Takes FASTQ as input:
   
• Output format: html and plain text
  
   

Don't run this code, it's FYI

FASTQC report example

• Example html output:
  (see https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ for more examples)

Genome index build example

(Manual: http://bowtie-bio.sourceforge.net/bowtie2/manual.shtml)

• Required for fast search of query sequence during mapping
• Takes fasta with reference as input (comma-separated list) and prefix name.
• Example run and resulting database:

Don't run this code, it's FYI

Read mapping example

• Takes query FASTQ and indexed database as input.
• Output is in SAM or BAM format
  (SAM format specification: https://samtools.github.io/hts-specs/SAMv1.pdf)
• SAM is a TSV file with header, BAM is its compressed binary version. The columns are:

Read mapping example

FASTQ

+ indexed database

-> SAM file

Samtools example

(Manual: http://www.htslib.org/doc/samtools.html)

• A toolkit designed for SAM/BAM manipulations
• View:
   
• Statistics assessment:
   
• Format conversion:
   
• Filtering:
  
  
  
  
  
   

Bedtools example

(Manual: https://bedtools.readthedocs.io/en/latest/content/bedtools-suite.html
BED format specification: https://genome.ucsc.edu/FAQ/FAQformat.html)

• A toolkit for conversion and manipulations with BED:
   
• A powerful tool for genomic arithmetics:
  (Tutorial: https://bedtools.readthedocs.io/en/latest/)

genomecov

map

intersect

Other formats

1. BED
   
   
   
    
2. BEDGRAPH
   
   
   
   
    
3. BIGWIG: binary compressed format, created with UCSC tools (see below)

UCSC genome browser & UCSC tools

• UCSC maintains genome browser and has a set of tools: https://genome-euro.ucsc.edu/

UCSC tools example

Full set of compiled tools: http://hgdownload.soe.ucsc.edu/admin/exe/linux.x86_64/

• We will need file conversion, check the documentation:

BEDGRAPH tracks visualization

We will use UCSC browser for bedgraph files visualization.

• Download both files file.genomecov.bedgraph and file.binned.bedgraph to your local computer.
   
• Manually add the header to your file:
  (see UCSC recommendations: https://genome.ucsc.edu/goldenPath/help/bedgraph.html)
  track type=bedGraph visibility=full
   
• Go to UCSC genome browser, hg38 genome:
  https://genome-euro.ucsc.edu/cgi-bin/hgTracks?db=hg38

UCSC genome browser

• Find "add custom tracks button".
• Check the upload form:

• Upload your modified BEDGRAPH files (with headers) to UCSC ("Choose file" -> "Submit").