Analysis of chromatin with ChIP-seq

Describe the process of ChIP seq

• Cross linking agent is added to cells - this fixes proteins of interest physically in place on the genome

• The genome is then fragmented into smaller fragments

  • this could be via sonication

• Antibodies against the protein of interest / specific histone modification are used to isolate specific regions of chromatin that are occupied by the studied protein

• The DNA is then purified and prepared into a library

• this is then submitted to massively parallel sequencing

• the sequence output is then aligned to a genome, with the counts used as a proxy of enrichment

• this determines where the POI is physically associated in the genome

How are libraries made for chip sequencing?

• In chIP, need to deal with fragmented samples and fix these to the slides

• Fragmentation via sonication produces very messy samples - have messy ends. these need to be converted into blunt ends before adaptor ligation

  • can use a polymerase with exonuclease activity (to chew back) or a polymerase activity to extend

  • this produces a blunt ended DNA with a clean 5’ and 3’ end

• Adaptors need to be added on to enable slide attachment for sequencing - blunt ended fragments are poor substrates for ligation so need to add a base on to make disjointed ends.

  • we can add a template independent polymerase, such a taq polymerase - this adds a single base overhang onto the blunted DNA , such as a single adenenine

• Adaptor sequences can then hybridise as they have a thymine overhang - enabling hybridisation and ligation to the adenine overhang on the sample

• Can then add a ligase, creating a covalent link between the adaptor and sample DNA

• We can then am[plify samples using a polymerase and oligonucleotide primers specific to the adaptor ends, this creates fragments with unique 5’ and 3’ ends that bind the flow cell

Describe the adaptor sequences used in library preparation for ChIP sq

• Adaptor sequences: hybridisation of of two fragments with a shared complementary seqeunce, as well as 2 non-complementary regions.

• will also have a thymidine overhang - this enables hybridisation and ligation to the adenine overhang on the sample

What data is produced by a massively parallel sequencer

• Data is in the form of DNA sequence

  • this can be produced in different formats for illumina - either single end or paired end

• All sequencers measure different parameters to determine the sequence, but each has a built in software codetermine the base and the certainty of this basecalling

  • this is used to produce data in a standardised format

what is the standardised format for sequencer output?

FASTQ file

What information is encoded in each line of a FASTQ file?

• line1 contains multiple bits of information

  • Sequencer name and run number

  • Position of cluster on the flow cell

  • read index

• sequence in the 5’-3’ direction

• space

• quality scores at each base

what is the read index in a fastq file useful for?

• can run many different sampels on one flow cell

• can tag each - index allows demultiplexiing of complex experiments

what does the quality scores at each base account for in a fastq file?

• probability of correctness in basecalling

• accounts for noise in seqeuneces

• this is encoded in an ascII key

  • this encodes numeric information

• Within scores, there is a certainty of basecalling accuracy

what can cause the quality of sequencing to differ?

• there is variability on the quality of sequencing depending on a few factors

  • density of clustering of DNA on a flow cell

    • undercluster → little data

    • over cluster → too close togehter, cant easily distinguish them during imaging

• need them to be separated enough to discern individual clusters, but dense enough to produce a high number of reads

how can programmes be used to improve sequencing quality?

• can plot the base position, quality score and plot the quality

• for each cycle you do → increased error rate

• can use programmes to rescue the useful regions!

  • these trim back reads until you reach a region where data is of good uality

What needs to be done once you have sequencing reads?

• Once you have the reads, they need to be aligned back to the genome to count the number of reads corresponding to agiven position of the genome

What is a common aligner for chIP-seq data? why is this useful?

• bowtie2 is a common aligner for chip0seq type aligner

  • gapped aligner - deals with small deletions and insertions in the data

  • also has statistical models to give information about the alignment qulity

what does bowtie2 do?

bowtie2 takes the reference genome and makes a matrix to compare the seqeunce reads to determine where they are in the genome → this speeds up the identification of where sequences are in the genome

how do we produce an alignment via bowtie2?

we feed the bowtie index and raw sequencing data into Bowtie2 → outputs an alignment file

what is the format of a bowtie2 alignment file?

this is in the format of a sequence alignment map - SAM

what do we convert the alignment file (SAM format) into?

we convert this into a binarised version, a BAM file

What information does a BAM file contain?

• fastq information

• additional information about the alignment:

  • read identifier

  • flag

  • chromosome number and positon

  • MAPQ

  • CIGAR

  • sequence and quality scores from fastq

  • optional fields - detail alignment quality

What does the flag in a BAM file denote?

• (whether the aligned sequence is forward or reverse - denoted by 0 or 16)

what does the MAPQ score in a BAM denote?

• this tells you how good the alignemnt is compared to the next possible alignment

• useful for gapped alignments and repetitive sequences

• 0-42

• if very low - lacks confidence → likley competitive!!

what does CIGAR in the BAM file denote?

tells us how many bases in the read undergo a full match with the underlying seeunce

Why do we need to visualise aligned chIP seq data?

• Need to visualise where seqeunces actually are in the genome

  • peaks indicate protein / modification enrichment

what would a Bam file input into a genome browser produce?

• If BAM file was put into a genome browser → little bars correlate to individual reads with seqeunce information

  • reads begin to pile up

  • BAM files are difficult to deal with, because they retain seqeunce level information etc - need less intense memory usage

What do we convert a BAM file into for visual analysis of chIP seq data? how do we do this?

convert into a bigWig file - using bam coverage

what does Bam coverage do?

• takes regions of the genome, sums and produces a value for the number of reads at a given position in the genome

  • produces a compressed binary file - bigwig

• turns individual seqeunce information bars into a single value of a cetain height

  • compressed and indexed in a binary format

How are bigWig files structured? what does this produce in IgV?

• BigWig files are structured to allow computer to isolate individual important pieces of the genome

• produces a linear region of the genome with underlying features - eg. genes

• can visualise and understand where a protein is distibuted across the genome

What information does the UCSC genome browser contain?

lots of information built in to allow comparisons and extraction of positions of genes and other information for comparative analysis

what information does the table browser contain?

extract where genes are → extract integral information to analyse the chIP-seq data

what is the format of the information for the UCSC genome browser

Bed file

What do we need to do to interrogate chip-seq data at the genome scale

• Need to take aligned information and combine it with bed files

• deep tools are useful for this

what are the inputs to use deeptools to combine data for genome wide analysis of chip-seq data?

• Inputs

  • USCS browser → table extraction component → isolated bed file corresponding to promoters (info about start and end)

  • bam coverage used to create bigWigs from raw alignment files

what does deep tools allow us to do?

deep tools allows us to integrate the intervals and the seqeuncing data

what does deeptools do with the inputs?

• creates a matrix with readcounts inherent to one data across all of the integrals → matrix file

  • this matrix file can be queried with other programmes to plot and quantitate the information

  • takes genome wide information and concentrates it into a format to only interrogate the regions you are interested in

once we have used deeptools, how can we quantitate genome data?

• Can then plot and quantitate data over the bed file regions

• could use R to do this

• Galaxy has programmes built in for visualisation of the data

• May want to metaplot the data or produce heat maps of the information across promoters

what do metaplots do? (promoter example)

• takes all promoters on average and the reigon surrounding

• plots the average intensity of sequencing across the elements

• example: produces a peak of RNA polymerase II downstream of the TSS

why are meta-plots deceiving?

• deceiving - this is a mean signal across regions - doesnt tell you about individual gene promoters

what do heat maps do? (promoter example)

• takes all promoters and aligns them on top of eachother, centred on the TSS

  • plots the intensity of these reads

• signal is not uniform across all poromoers

• makes sense!! - not every gene is expressed in every cell type

How can we identify bound regions of the genome de novo for regions that have not been ChIP-ped before?

• often use MACS2 programme

  • uses the bowtie alignment and applies a statistical approach to identify where regions of the enrichment are via cross correlation analysis

• cross correlation analysis - looks for a distribution of reads pointing in a direction consistent with a statistical model

• Can also use different parameters to drive how sensitive it is in detection
  → takes information and produces a bed file → tells you where the peaks are in the genome

How can we use software to interrogate enrichments?

• Many different software tools used to interrogate where enrichments might be

• can do simple queries at the genome scale to compare different bed files

  • eg. protein may bind in different peaks across the genome

  • bed tools use simplistic binary comparisons of these

  • can identified shared and unique regions and query multiple factors

• can do simple bed analysis - is it there or not?

• can also take the bed intervals → extract underlying sequence from bed regions → carry out motif analysis to determine whether there are specific sequences associated wiyh binding

  • may indicate DNA binding specificity

How can interval analysis be useful?

• useful for more detailed quantitative analysis - eg. compare the relationship between two parameters

• this can be quantitated - carry out statistical analysis

how can genomic approaches be used to study system perturbation changes?

• can do knockouts and compare signals between wt and knockout cells via chip analysis

what software can be used to study system perturbations using chip?

• diff bind can be used

  • take bowtie alignments and bam file → read into MACS2 - unbiasedly calls where intervals of enrichment is in the data sets

  • apply raw seqeunding data from the bam file and peaks → creates a tabular file of the underlying statistical analysis

  • needs repeat!!

• takes interval information across the conditions → uses the replicated data to produe a log2 fold change

  • gives p value of false discovery rate

what does an MA plot produce?

• Can generate an MA plot from the tables

• plots the starting intensity of reads at an interval and the log2 fold change

  • below - goes down

  • shows that most of the sites identified are going down

• can also try and determine whether this decrease is uniform

  • can compare in the perturbed and unperturbed stat