Chromatin Immunoprecipitation DNA Sequencing (ChIP-seq)

1
Chromatin Immunoprecipitation DNA Sequencing
(ChIP-seq)
2
2nd and 3rd Generation DNA Sequencers and
Applications

• Roche 454 (2nd)
• Illumina Solexa(2nd)
• ABI SoLid (2nd)
• Helicos (3rd)

• Applications
• De novo sequencing
• Targeted resequencing
• Digital Gene Expression (DGE)
• RNA-seq
• ChIP-seq

Sequencing Platforms
3
Why ChIP-seq?

• Protein-DNA interactions
• Chromatin States
• Transciptional regulation

4
ChIP experiment

• In Nutshell
• Protein cross-linked to DNA in vivo by treating
  cells with formaldehyde
• Shear chromatin (sonication)
• IP with specific antibody
• Reverse cross-links, purify DNA
• PCR amplification
• Identify sequences
• Genome-wide association map

-unless using a single molecule sequencer
5
History From ChIP-chip to ChIP-seq

• ChIP-chip (c.2000)
• Resolution (30-100bp)
• Coverage limited by sequences on the array
• Cross-hybridization between probes and
  non-specific targets creates background noise

6
ChIP-seq experiment (2007-present)
7
Sample Prep Solexa vs. Helicos
8
ChIP-seq Materialsample preps with in-house
protocols
Helicos sample prep Normal QC and ChIP
steps Input material 3ng-9ng RNAseA/ProteinaseK
treatment (2-3h) Purification (phenol/precipitatio
n) (1.5h) Tailing (1.5h) Termination
(1.5h) Amount of library sequenced
approx. 1/3 Unique Tags after analysis approx
gt12M (based on our limited ERaChIP-seq
libraries) Slide borrowed from Thomas
Westerling
Solexa sample prep Normal QC and ChIP
steps Input material typically gt30ng End-Repair
(1h) Purification (phenol/precipitation) (1.5h)
A-overhang (1h) Purification (phenol/precipitatio
n) (1.5h) Adapter oligo ligation
(30min) Purification (phenol/precipitation)
(1.5h) Size-selection (30min by
E-gel) Precipitation (1h) Amplification PCR (2h)
(12-18 cycles) Size-selection (30min by
E-gel) Precipitation (1h) Diagnostic gel (30min)
QC by direct qPCR (4hours) Amount of library
sequenced approx. 1/10 Unique Tags after
analysis gt 3M (based on our limited ERaChIP-seq
libraries)
9
(No Transcript)
10
Helicos vs Solexa vs ChIP2
Solexa data (red) Unique tags 4M Peaks called 10
500A Negative peaks 20 000B
2900
1. Solexa
2541
433
Helicos data (blue) Unique tags 13M Peaks called
12 500 Negative peaks 1000E
4700
2. Helicos
3. ChIP2
3744
5293
ChIP2C data (green) Array technology, no
tags Peaks called 12 500 FDR 20D
1661
A) More inclusive (10) ELAND mapping used
(compare to Bowtie in library table) B) MACS
performs a sample swap between ChIP and Input
(chromatin) samples and calculates a local
?-value to determine level of background peaks
called in control data. This gives a FDR for each
positive peak. Due to the nature of deep
sequencing combined with PCR this parameter is in
some sample extremely high and not entirely
trustworthy. C) ChIP2 data published in
Carroll et al. Nat Genet. 2006 Nov38(11)1289-97.
D) FDR values of ChIP2 are calculated
differently from FDRs by MACS and are not
directly comparable. E) Negative peaks and thus
local FDR values are at first glance more
reliable in Helicos sequencing, in part at least
due to the lack of amplification the removes
scientist introduced artifacts and reduced
complexity of sequenced library.
11
ChIP-seq Analysis
12
ChIP-seq peaks

• Only 5 end of fragments are sequenced
• Tags from both and - strand aligned to
  reference genome

13
/- tag mapping
14
Types of Analysis

1. Binding site identification and discovery of
   binding sequence motifs (Non-histone ChIP)
2. Epigenomic gene regulation and chromatin
   structure (Histone ChIP)

15
Binding Site DetectionBut where does the meat go?
16
Control Input DNAMeasuring enrichment
Input DNA portion of DNA sample removed before IP
Rozowsky, J. et al. PeakSeq enables systematic
scoring of ChIPSeq experiments relative to
controls. Nature Biotech. 27, 66-75 (2009)
17
Why we need to sequence Input DNA

• Input DNA does not demostrate flat or random
  (Poisson) distribution
• Open chromatin regions tend to be fragmented
  more easily during shearing
• Amplification bias
• Mapping artifacts-increased coverage of more
  mappable regions (which also tend to be
  promotor regions) and repetitive regions due
  inaccuracies in number of copies in assembled
  genome

18
Depth of SequencingAre we there yet?
19
ERa E2 Helicos MACS peaks 12500(tag30 mfold30)
sequence depth determination by subsampling
peaks detected of total peaks/bin
of tags sampled
FoldChange Bins 0-20 20-40
40-60 60-80 80-100 100-120 120-140 140-160 160-180
180-200 200-220 Number of total 7687 2841
935 429 217 140 85 49 23
7 4 Peaks in each bin
20
Statistical Significance
21
MACS shifted tag-count graph i.e. Peak shapes
Helicos Input
HelicosChIP
SolexaChIP
Solexa Input
22
MACS shifted tag-count graph i.e. Peak shapes
Helicos Input
HelicosChIP
SolexaChIP
Solexa Input
23
MACS shifted tag-count graph i.e. Peak shapes
Helicos Input
HelicosChIP
SolexaChIP
Solexa Input