Week 2: ChIPseq

Section Links

Week 2 Overview

For week 2, you will be performing a quick quality control check by plotting the correlation between the bigWigs you generated last week. Then, you will be performing standard peak calling analysis using MACS3, generating a single set of reproducible and filtered peaks, and annotating those peaks to their nearest genomic feature.

Objectives

Plot the correlation between the bigWig representations of your samples
Perform peak calling using MACS3 on each of the two replicate experiments
Use bedtools to generate a single set of reproducible peaks with ENCODE blacklist regions filtered out
Annotate your filtered, reproducible peaks using HOMER

Containers for Project 2

FastQC: ghcr.io/bf528/fastqc:latest

multiQC: ghcr.io/bf528/multiqc:latest

bowtie2: ghcr.io/bf528/bowtie2:latest

deeptools: ghcr.io/bf528/deeptools:latest

trimmomatic: ghcr.io/bf528/trimmomatic:latest

samtools: ghcr.io/bf528/samtools:latest

macs3: ghcr.io/bf528/macs3:latest

bedtools: ghcr.io/bf528/bedtools:latest

homer: ghcr.io/bf528/homer:latest

Plotting correlation between bigWigs

Recall that the bigWigs we generate represent the count of reads falling into various genomic bins of a fixed size quantified from the alignments of each sample. Assuming the experiment was successful, we naively expect that the IP samples should be highly similar to each other as they should be capturing the same binding sites for the factor of interest. Following this logic, the input controls which represent a random background of DNA from the genome should be different from the IP samples and similar to each other.

We are going to perform a quick correlation analysis between the distances in our bigWig representations of our BAM files to determine the similarity between our samples with the above assumptions in mind.

Create a module and use the multiBigwigsummary utility in deeptools to create a matrix containing the information from the bigWig files of all of your samples.
Create a module and use the plotCorrelation utility in deeptools to generate a plot of the distances between correlation coefficients for all of your samples

Peak calling using MACS3

In plain terms, peak calling algorithms attempt to find areas of enriched reads in a genome relative to background noise. MACS3 (Model-Based Analysis of Chip-Seq) is a commonly used tool that incorporates a Poisson model and other innovations to make robust peak-finding predictions.

Use the MACS3 manual for the callpeak utility and create module that successfully runs callpeak
Ensure that you specify the -g flag correctly for the human reference genome and the -f flag explicitly for paired-end BAM files.
You will need to figure out how to pass both the IP and the Control sample for each replicate to the same command. i.e. callpeak should run twice (IP_rep1 and control_rep1) and (IP_rep2 and control_rep2) as ChIP-seq experiments have paired IP and controls.

Generating a set of reproducible peaks with bedtools intersect

We discussed various strategies for determining a set of “reproducible” peaks. For the sake of expedience, we will be performing a simple intersection to come up with a single set of peaks from this experiment. Please come up with a valid intersection strategy for determine a reproducible peak. Remember that this choice is subjective, so make a choice and justify it

Use the bedtools intersect tool to produce a single set of reproducible peaks from both of your replicate experiments

Filtering peaks found in ENCODE blacklist regions

In next generation sequencing experiments, there are certain regions of the genome that have been empirically determined to be present at a high level independent of cell line or experiment. These unstructured and anomalous regions are problematic for certain analyses (ChIPseq) and are considered to be signal-artifact regions and commonly stored in the form of a blacklist

The Boyle LAB as part of the ENCODE project have very kindly produced a list of these regions in some of the major model organisms using a standard methodology. This list is encoded as a BED file and is hosted by the Boyle Lab/ Please copy the provided file from references/ to your repository.

Create a module that uses bedtools to remove any peaks that overlap with the blacklist BED for the most recent human reference genome.
Typically, any peaks that overlap a blacklisted region by even 1bp will be removed. You may choose a different strategy if you prefer as long as you justify your choice in the write-up later.

Annotating peaks to their nearest genomic feature using HOMER

Now that we have a single set of reproducible peaks with signal-artifact blacklisted regions removed, we are going to annotate our peaks by assigning them an identity based on their closest genomic feature. While we have discussed many caveats to annotating peaks in this fashion, it is a quick and exploratory analysis that enables quick determination of the genomic structures your peaks are located in and their potential regulatory functions.

Create a module that uses homer and the annotatePeaks.pl script to annotate your BED file of reproducible peaks (filtered to remove blacklisted regions).

0.2 Week 2 Tasks Summary

Create nextflow modules and a script that performs the following tasks:
1. Create a correlation plot between the sample bigWigs using deeptools multiBigWigSummary and plotCorrelation
2. Use MACS3 callpeak to perform peak calling on both replicate experiments
3. Generate a single set of reproducible peaks using bedtools
4. Filter peaks contained within the ENCODE blacklist using bedtools
5. Annotate peaks to their nearest genomic feature using HOMER

Week 1: ChIPseq

Week 3: ChIPseq