Download deschamp_2009_sequencing

Overview and Applications of
Next-Generation Sequencing
Technologies
Stéphane Deschamps
Analytical & Genomic Technologies
DuPont Agriculture & Nutrition
Pioneer Hi-Bred International
Outline
1. Next-Generation Sequencing Platforms
1. 454 FLX technology
2. Solexa/Illumina technology
2. Applications of Next-Generation Sequencing Technologies
1. Overview
2. Variant detection with Illumina platform
3. Open-source tools for bioinformatics
3. Third-Generation Sequencing technologies: what’s next?
Sanger sequencing
Successive improvements now allows 96 800-900 base reads to be sequenced in less than 2h
Sanger sequencing
Sanger sequencing has been, and still is, very useful...
...but it remains slow and expensive
Sequencing Platform Comparisons
ABI
3730xl
454 FLX
Titanium
Illumina
GA II
Read Length
~750bps
~450bps
18-75bps
Number of
reads/run
96
500K
100MM
Max Yield/run
~70Kbps
~1Gbp
~10Gbps
Cost/1Gbp
$3.5MM
$7,000
$1,000
Run time/machine
to 1Gbp
8 years
1 day
<1 day
Next-Generation Sequencing
Second-generation platforms:
Third-generation platforms:
•454/Roche
•Solexa/Illumina
•SOLiD/ABI
•Helicos BioSciences
•Dover Systems
•Complete Genomics
•BioNanomatrix
•VisiGen
•Pacific Biosciences
•Intelligent Bio-Systems
•ZS Genetics
•Reveo
•LightSpeed Genomics
•NABsys
•Oxford Nanopore Technologies
454 FLX Titanium
• First next-generation sequencing platform launched (October
2005)
• Titanium chemistry for the 454 FLX launched in September 2008
• Sequencing By Synthesis
– Pyrosequencing
– Chemiluminescent signal
• Long read technology (~450 nucleotides)
• Possibility of sequencing both ends of
DNA fragments (FLX platform)
• Generates up to 0.5Gbps per run
• Max cost is ~$10,000/run
454 FLX Titanium
• DNA Library Construction
• Emulsion PCR
• Sequencing
DNA Library Construction
• DNA fragmentation via nebulization
• Size-selection
• Ligation of adapters A & B
• Selection of A/B fragments via biotin selection
• Denaturation to select single-stranded A/B fragments
• No cloning!
(B/B)
A/B ss DNA
End repair
Streptavidin
Denaturation
Streptavidin
+
(A/B)
+
(A/A)
Emulsion PCR
Emulsion PCR
•
Add DNA to capture beads (needs titration)
•
Add PCR reagents to DNA and capture beads
•
Transfer sample to oil tube or cup
•
Emulsify DNA capture beads in PCR reagents
to form water-in-oil “microreactors”
– Emulsion with Qiagen TissueLyser (highspeed shaker)
•
Clonal amplification in microreactors
– Careful not to break the emulsion!
– ~10MM copies per capture bead
•
Break emulsion and enrich for DNA positive
beads
– Use biotinylated oligo to capture enriched
beads then denature
www.roche-applied-science.com
Bead deposition into plates
• Deposition of enriched beads into
PicoTiter plate
• Well diameter = 29uM allowing for a
single bead (20uM diameter) per well
• Chambers are filled with enzyme
beads, DNA beads and packing
beads.
www.roche-applied-science.com
Pyrosequencing
1.
Polymerase add nucleotide
(sequential flow of dNTPs)
2.
PPi is released
3.
Sulfurylase creates ATP
from PPi
4.
Luciferase hydrolyzes ATP
and use luciferin to make
light
www.roche-applied-science.com
Image and signal processing
1. Raw data is series of
images (one image per
base per cycle).
2. Data are extracted,
quantified and
normalized.
3. Read data are converted
into “flowgrams”.
Post-processing
1.
Output = flowgrams, basecalls, Phred-equivalent scores
2.
Basecall & Flowgrams can be used in the following applications:
1.
De novo assembler – consensus sequences assembled into
contigs with quality scores and ACE file (works best with genomic
DNA).
2.
Reference mapper – contigs mapped to reference sequence + list
of high-confidence mutations
3.
Amplicon variant analyzer – identification of sequence variants in
amplicon libraries
Illumina Genome Analyzer
• Successor to MPSS (Massively Parallel Signature Sequencing)
• Single molecule array (“flow cell”) with millions of amplified
clusters
• Sequencing By Synthesis
– Removable fluorescence
– Reversible terminators
• Short read technology (16 - 75 nucleotides)
• Possibility of sequencing both ends of DNA fragments
• Generates up to 20Gbps per run
• Max cost is ~$10,000/run
= $500/Gbp!
Illumina Genome Analyzer
Sample Prep
Prepare DNA
fragments
+
Ligate
adapters
Cluster Station
Cluster Synthesis
Genome Analyzer
Sequencing
Analysis Pipeline
Cluster Station
Genome Analyzer
Fluidics and
Electronics
Flow Cell &
Detection
Laser
Optics
Cluster Generation
or RNA
- anneal
Cluster Generation
- extension
DNA Clusters
• ~1,000 copies of DNA in each cluster
• 1-2 microns in diameter
Reversible Terminator Chemistry
Sequencing by Synthesis (SBS)
5’
3’
Cycle 1:
First base incorporated
G
T
Remove unincorporated bases
A
C
A
T
G
C
C
G
T
T
A
C
A
Add sequencing reagents
C
G
A
T
T
A
G
A
C
T
C
C
G
A
G
C
T
C
G
A
5’
T
Detect signal
Deblock (removal of fluorescent dye
and protecting group)
Cycle 2 - n: Add sequencing reagents and repeat
Sequencing by Synthesis (SBS)
Data Analysis Workflow - Illumina
Images (.tif)
Illumina Analysis Pipeline
Data
transfer
and
Storage
Image
Analysis
Base
calling
Sequence
Analysis
alignment (ELAND),
filtering (chastity)
•Cluster Intensities
•Cluster Noise
•Cluster Sequence
•Cluster Probabilities (Scores)
•Corrected Cluster Intensities
• cross-talk correction
• phasing correction
1 image per dye
4 dyes/cycle
75 cycles
50 tiles/column
2 columns/lane
8 lanes/flowcell
240,000 images
per flowcell
x8 MB per image
1.92 TB of image data
•Image analysis module is Firecrest
x2 for PE run
3.8 TB of image data
•Base calling module is Bustard
•Sequence analysis module is Gerald
Alignments,
Assemblies, Normalization,
Annotations &
Post-processing Evaluations
Other platforms
Sequencing
Sequencing
Run
Read
Reads per
Throughput per
Platform
Chemistry
Time
Length (bp)
Run (million)
Run (Gbp)
Roche 454 FLX
Pyrosequencing
10h
400-500
~1
0.4-0.5
9.5 days
100
250
25
8 days
50
400
20
8 days
25-55
600-800
15-45
80h
28
300-400
10
Sequencing by
Synthesis
Sequencing by
ABI SOLiD
Ligation
Sequencing by
Helicos HeliScope
Synthesis
Sequencing by
Polonator
Ligation
Illumina GAIIx
Data Storage & Quality
Images?
~Phred 20
Phred score 20 = 1% error rate
Quality vs. Read Length? Trimming?
Lower sequence quality than Sanger sequencing but offset by deeper coverage
Single short read uniqueness
Illumina 35 base reads aligned to A. thaliana genome
10000000
~4MM reads
1 location
2 locations
3 locations
4 locations
5 locations
6 locations
7 locations
8 locations
9 locations
10 locations
100000
10000
1000
100
10
1
35
34
33
32
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
Number of Reads
1000000
Read length (bps)
Applications of
Next-Generation Sequencing
Gene Expression Profiling
– Tag count & Alignments
– Digital Gene Expression Tag Profiling
• Short cDNA fragments mapping to 3’ ends of transcripts
• SAGE-like approach (1 short tag/transcript)
• 20 base tag output (RE site + 16 bases) aligned to a reference genome
• Identify, quantify and annotate expressed genes
– Transcriptome Profiling (RNA-Seq)
• cDNA fragments generated via random priming
• 36-75 base output aligned to a reference genome
• Assemble entire transcript sequence
• Identify, quantify and annotate expressed genes
• Identify SNPs, alleles and alternative splice variants
Tag Profiling – Sample Prep (Illumina)
Total RNA (5ug)
mRNA isolation
AAAAA
1st and 2nd Strand cDNA Synthesis
AAAAA
TTTTT-bio
Restriction Enzyme Digestion (DpnII or NlaIII)
AAAAA
TTTTT-bio
CATG
GEX Adaptor 1 Ligation
MmeI
CATG
GTAC
AAAAA
TTTTT_bio
MmeI
CATG
GTAC
MmeI digestion
NN
GEX Adaptor 2 Ligation
NN
NN
CATG
GTAC
PCR Amplification
PCR Primer 2
PCR Primer 1
CATG
GTAC
sequencing primer
TAG
Cluster
Generation
Transcriptome Profiling – Sample Prep (Illumina)
Tissue
Total RNA isolation (10ug)
mRNA isolation
AAAAA
Fragmentation (random)
AAAAA
1st and 2nd Strand cDNA Synthesis (N6 primer)
AAAAA
TTTTT
Adaptor Ligations
PCR Amplification
PCR Primer 2
PCR Primer 1
sequencing primer 1
sequencing primer 2
Cluster
Generation
Novel Transcript Discovery
–
Small RNA Identification and Profiling
•
–
Small RNA size is suitable to discovery with next-generation sequencing
Deep assessment of alternative splicing isoforms
•
Deep coverage allows discovery of rare isoforms
Mortazavi et al. (2008), Nat. Methods
De novo Sequencing
– Whole Genome Sequencing
• Small genomes that are not too complex (microbial)
• The longer the reads, the better – 454 chemistry most suitable
• Paired-End sequencing
– Whole Transcriptome Sequencing
– Targeted Sequencing
• Pooled PCR products
– Raindance Technologies (~4,000 amplicons in one tube)
– Padlock probes
• Pooled BAC clones
• Sequence Capture (Solid phase, Liquid phase)
– Agilent, Febit & Nimblegen
– Metagenomics & Microbial diversity
Gene Regulation
–
ChIP-Seq (immunoprecipitate sequencing)
•
Capture regions of the genome bound by proteins (transcription factors,
histones)
•
Sequences need to be aligned to a reference sequence
•
Requires complex algorithm to determine differential levels of coverage
throughout the genome
–
Methyl-Seq (methylation status) – Bisulfite Sequencing
•
–
Sequences aligned and compared to reference sequence
DNAseI Hypersensitivity Site Sequencing
Mikkelsen et al. (2007), Nature
Variant & Structural Variation
– Coverage & Alignment
– Paired-End Sequencing
– Whole Genome Resequencing
• Small genomes that are not too complex (repeats, duplications...)
• The longer the reads, the better
– Targeted Resequencing
• Complex genomes (crops)
– Reduced representation libraries (methyl-sensitive enzymes)
– Transcriptome
• Sequence Capture (Microarrays)
» Agilent, Febit & Nimblegen
» CGH arrays
Challenges in variant discovery
1.
Base quality & filtering (scoring threshold)
2.
Sequencing errors vs. SNPs
1.
To differentiate true polymorphisms from sequencing errors
2.
Coverage of a given SNP region and redundancy of reads (coverage vs. number of samples)
3.
Availability of a reference sequence (genome)
1.
To separate unique vs. duplicated sequences
2.
Duplication in one line but not another
3.
Polymorphism rate in one line vs. another = need to set conditions for alignment
4.
Paired-end sequencing can help unique read placement
5.
Complex genomes = need to reduce complexity prior to sequencing
1.
High repeat content (ex: ~80% in maize, ~70% in soy, 90% in sunflower…)
2.
Gene duplications and genome plasticity (polyploidy, partial or whole genome duplications...)
Reduced-representation libraries
1.
DNA methylation in plants occurs at 5-methyl cytosine within CpG dinucleotides and
CpNpG trinucleotides
2.
Transposons and other repeats comprise the largest fraction of methylated DNA. Studies in
Arabidopsis have shown that CG sites in the 3’ end of the transcribed regions of more than
one third of all genes also are methylated (Zhang, Science, 320, 489, 2008).
3.
Methylation is critically important in silencing transposons and regulating plant development
(methylation in promoters appears to reduce transcription)
PstI sites
transposon
transposon
transposon
PstI digestion
Recover digested fraction (gel, column)
P
P
P
PP
P
P
P
Library Construction
Genomic DNA
Digestion with one methyl-sensitive restriction enzyme (RE) and fractionation
Ligation of biotinylated RE-specific adapters 1
B
B
Digestion with 4-bp cutter (DpnII)
GATC
B
Ligation of DpnII-specific adapter
GATC
CTAG
B
Binding to streptavidin column and digestion with RE
GATC
CTAG
Ligation of RE-specific adapters 2
GATC
CTAG
PCR enrichment, gel purification, size selection (150-500bp fragments),
cluster synthesis and sequencing (36 cycles)
Deschamps et al. The Plant Genome (in press)
SNP detection flowchart
Basecalling, cropping last 4 bases & initial base-quality filter (for individual tags)
Filtering and
Condensing
Condensing & optional consensus base-quality filter (for unitags sequences)
Creating HQ unitag datasets (removing singlets)
Comparing HQ unitag datasets from genotype “A” and genotype “B” using Vmatch
Comparing
two genotypes
Filtering, to accept clusters with only two members (A, B) with exactly one mismatch
Recovering matched HQ unitag sequences and SNP sites from Vmatch alignments
Mapping to
genome
Mapping SNP-containing HQ unitags to reference sequence (genome),
using a k-mer table (k=length of trimmed tags), and find copy numbers and locations.
Capturing single-copy HQ unitags with up to a single-base mismatch to the
reference sequence at the exact location of the putative SNP site for one or both genotypes.
Example: one flow cell in soybean (Williams82 vs. Pintado)
Run Metrics
Williams82
Pintado
Number of total reads generated
(after initial basecalling)
37,666,279
38,000,474
Number of filtered total reads †
24,519,484
23,101,973
Number of unitags (generated
from filtered total reads)
965,610
885,429
Number of high quality (HQ)
unitags ‡
255,918
246,102
Scatter Plot
100,000
Frequency
Alignment of HQ unitags
against the reference sequence:
10,000
1,000
100
10
Zero mismatch §
208,923
197,015
One mismatch §
27,770
27,699
Two or more mismatches
§
1
10
19,225
21,388
152,185
144,559
100
log10(Depth)
1,000
10,000
Depth
HQ unitags aligning uniquely to
the reference sequence with zero
mismatch
† Filtered
total reads defined as having a quality value for individual base greater than or equal to 15
‡
HQ unitag reads defined as having a quality value for each base greater than or equal to 15, and with an individual read
count greater than or equal to 2.
§ Best
match to reference sequence of HQ unitag reads aligning uniquely or multiple times to the reference sequence
100,000
Results & Validation
Experiments
Q Score threshold: 15
Soy: Williams82 vs. Pintado
Rice: Kasalath vs. Taichung65
Q Score threshold: 25
Soy: Williams82 vs. Pintado
Rice: Kasalath vs. Taichung65
*
Not Confirmed
*
Putative SNPs
Confirmed
Validation rate
1,682
2,618
163
162
5
6
97.0%
96.4%
702
2,148
168
174
2
1
98.8%
99.4%
*SNPs confirmed/not confirmed via Sanger sequencing of PCR products for both genotypes
Distribution of HQ unitags & SNPs related to annotated gene density (soybean)
Gene Density (excluding TEs) in 500Kb window
Coverage by HQ unitags in 70Kb window
SNP Density in 70Kb window
Distribution of HQ unitags & SNPs related to distance to
annotated genes (excluding TEs) in soybean
38
.3
50.0
.6
21
.4
SNPs
18
.3
20
16
.2
20.0
24
.4
23
.8
22
.8
21
.9
24
.1
30.0
34
.5
33
.7
40.0
Pintado
Williams82
10.0
0.0
intron
CDS
within_5k
intergenic
Intron, CDS and
UTR coordinates
determined from
GFF annotation
files
Bioinformatic tools
Alignment and Polymorphism Detection
1. SOAP – Short Oligonucleotide Alignment Program
•
Ruiqiang Li, Beijing Genomics Institute
•
http://soap.genomics.org.cn
2. MAQ – Mapping and Assembly with Quality
•
Heng Li, Sanger Centre
•
http://maq.sourceforge.net/maq-man.shtml
3. Bowtie - An ultrafast memory-efficient short read aligner
•
Ben Langmead and Cole Trapnell, University of Maryland
•
http://bowtie-bio.sourceforge.net/
4. ssahaSNP – Tool to detect homozygous SNPs and indels
•
Adam Spargo and Zemin Ning, Sanger Centre
•
http://www.sanger.ac.uk/Software/analysis/ssahaSNP
Bioinformatic tools
Genomic Assembly
1. Velvet – De novo assembly of short reads
•
Daniel Zerbino and Ewan Birney, EMBL-EBI
•
http://www.ebi.ac.uk/~zerbino/velvet/
2. SSAKE – Assembly of short reads
•
Rene Warren, et al, British Columbia Cancer Agency
•
http://bioinformatics.oxfordjournals.org/cgi/content/full/23/4/500
3. Euler – Genomic Assembly
•
Pavel Pevzner and Mark Chaisson, University of California, San Diego
•
http://nbcr.sdsc.edu/euler/
www.illumina.com
Bioinformatic tools
ChIP Sequencing
1. ChIP-Seq Peak Finder
•
Barbara Wold, Cal Tech and Rick Meyers, Stanford University
•
http://woldlab.caltech.edu/html/software/
Digital Gene Expression
1. Comparative Count Display
•
Alex Lash, NIH
•
ftp://ftp.ncbi.nlm.nih.gov/pub/sage/obsolete/bin/ccd/
2. SAGE DGED Tool
•
Cancer Genome Anatomy Project
•
http://cgap.nci.nih.gov/SAGE/SDGED_Wizard?METHOD=SS10,LS10&O
RG=Hs
www.illumina.com
Bioinformatic tools - Illumina
Overview
1. Obtain Bustard reads and align
against Genome with Eland
2. Aggregate and SNP call data with
CASAVA
3. GenomeStudio™ wizard import of
data
4. Examine coverage and quality in
stacked alignment graphs for a
selected region/chromosome
5. Export table of SNPs and
consensus sequence
Bioinformatic tools - Illumina
Third-Generation Sequencing
technologies: what’s next?
Next-Generation Sequencing
Second-generation platforms:
Third-generation platforms:
•454/Roche
•Solexa/Illumina
•SOLiD/ABI
•Helicos BioSciences
•Dover Systems
•Complete Genomics
•BioNanomatrix
•VisiGen
•Pacific Biosciences
•Intelligent Bio-Systems
•ZS Genetics
•Reveo
•LightSpeed Genomics
•NABsys
•Oxford Nanopore Technologies
Pacific Biosciences
• SMRT™ Technology (to be commercially
launched Fall 2010)
• Single DNA polymerase attached at bottom
surface of nanometer-scale hole, incorporates
in real-time fashion fluorescently labeled
nucleotides to elongated strand of DNA
• Elongated strand can be several thousands of
nucleotides in length
www.pacificbiosciences.com
Pacific Biosciences
1.
Small size of the hole favors rapid in-and-out diffusion of nucleotides and dye following
their cleavage. Meanwhile, incorporated nucleotide is held within the detection volume
for 10’s of milliseconds, order of magnitude longer than the time it takes for nucleotides
to diffuse in and out of the hole, therefore decreasing background noise
2.
Fluorescent dye is attached to the phosphate chain, rather than the base, and is
cleaved when the nucleotide is incorporated to the DNA strand.
=> Decreased background noise and use of phospholinked nucleotides circumvents the need
for successive cycles of incorporation, washing, scanning and removal of the label,
therefore optimizing processivity of the enzyme and allowing longer read lengths
=> No need for washing decreases the consumption of reagents
Nanopore Sequencing
= the real $100 genome?
1.
Sequencing-by-Synthesis requires lots of preparation, lots of reagents (polymerase,
nucleotides, fluorescent labels...) and expensive detection systems.
2.
Nanopore sequencing does not rely on amplification or labeling, and provides a direct
electrical signal for base calling. It is based on a simple idea of “passing” DNA
fragments through a nanometer-scale pore and detecting in a real-time fashion signal
due to the DNA blocking the electrical current that runs through the pore
3.
Oxford Nanopore: Protein nanopore
1.
Long read lengths (1000’s)
2.
High read accuracy
3.
Current technical issues:
•
Exonuclease
Alpha-hemolysin
Attachment of the exonuclease
to the pore
•
Parallelization
(1,000’s of pores per chips)
www.nanoporetech.com
Cyclodextrin
(encapsulate
nucleotide)
Questions?