Transcript PowerPoint Presentation - No Slide Title

SNP Discovery and Genotyping Workshop
• SNP discovery strategies
Debbie Nickerson
• Identifying SNPs by association for genotypephenotype analysis of candidate genes
Chris Carlson
• Identifying haplotypes for genotype-phenotype
analysis of candidate genes
Dana Crawford
• SNP genotyping strategies
Debbie Nickerson
SNP Discovery and Genotyping Strategies
Debbie Nickerson - [email protected]
• Overview of Variation in the Human Genome
• SNP Discovery Strategies and Status
• SNP Data in the PGAs
• Genotyping SNPs
Total sequence variation in humans
Population size:
6x109 (diploid)
Mutation rate:
2x10–8 per bp per generation
Expected “hits”:
240 for each bp
Every variant compatible with life exists in the population
BUT: Most are vanishingly rare
Compare 2 haploid genomes: 1 SNP per 1331 bp*
*The International SNP Map Working Group, Nature 409:928 - 933 (2001)
Strategies to Find SNPs
• Mine them from Existing Genome Resources
• Targeted SNP Discovery in Candidate Genes
Berkeley PGA - http://pga.lbl.gov
CardioGenomics - http://www.cardiogenomics.org
InnateImmunity - http://innateimmunity.net
SeattleSNPs - http://pga.mbt.washington.edu
Southwestern - http://pga.swmed.edu
Sequence-based SNP Mining
BAC library
BAC Overlap
Genomic DNA
mRNA
RRS Library
or Sampling
cDNA Library
Shot gun Overlap
EST Overlap
Sequence Overlap SNP discovery
GTTTAAATAATACTGATCA
GTTTAAATAATACTGATCA
GTTTAAATAGTACTGATCA
GTTTAAATAGTACTGATCA
~ 4.1 Million SNPs Available http://www.ncbi.nlm.gov/SNP/
Mining Finds Only A Small Fraction of the SNPs
Fraction of SNPs Discovered
A
1.0
96
48
G
24
16
8
0.5
2
0.0
0.0
0.1
0.2
0.3
Minor Allele Frequency
0.4
0.5
Total Estimated SNPs and Fraction in dbSNP
minimal allele expected SNPs expected SNP
frequency
(millions)
frequency (bp)
expected % in
database
1%
11.0
290
11-12
5%
7.1
450
15-17
10%
5.3
600
18-20
20%
3.3
960
21-25
30%
2.0
1570
23-27
40%
0.97
3280
24-28
L. Kruglyak and D. Nickerson, Nat Genet 27:234-236 2001
Surfactant B - Locus Link
dbSNP
(http://www.ncbi.nlm.nih.gov/SNP/)
Surfactant B - dbSNP
Confirmation of SNP Resource in New Sample
Potential Pitfalls
Confirmed Multiple Method Report in dbSNP
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Confirmed Unique Method Report in dbSNP
Strategies to Find SNPs
• Mine them from Existing Resources
• Targeted SNP Discovery in Candidate Genes
Berkeley PGA - http://pga.lbl.gov
CardioGenomics - http://www.cardiogenomics.org
InnateImmunity - http://innateimmunity.net
SeattleSNPs - http://pga.mbt.washington.edu
Southwestern - http://pga.swmed.edu
Sequence-based SNP Identification
Amplify DNA
5’
3’
Sequence
Phrap
Phred
Sequence each end
Base-calling
Contig assembly
of the fragment.
Quality determination
Final quality determination
PolyPhred
ATAGACG
ATAGACG
ATACACG
ATACACG
ATAGACG
ATACACG
Polymorphism detection
Consed
Sequence viewing
Polymorphism tagging
Analysis
Homozygotes
Heterozygote
Polymorphism reporting
Individual genotyping
Phylogenetic analysis
Sequence-Based Detection and Genotyping of SNPs
Jim Sloan, Tushar Bhangle (PolyPhred)
Matthew Stephens, Paul Scheet (Quality Scores for SNPs)
Phil Green, Brent Ewing, David Gordon (Phred, Phrap, Consed)
PGA SNPs
• The PGAs provide a validated SNP resource
(Allele Frequency Data)
• Novel Views of the Variation Data
 Emerging Pathway Interfaces
 Color Fasta Formats
 Gene Structure Views
 Visual Genotypes
 Linkage Disequilibrium Views
 TagSNPs
 Haplotypes
• Many New Formats Under Development
Toward comprehensive association studies
• 5-7 million common variants exist in genome
• Testing all for association is impractical today
• Can the list be reduced w/o loss of power?
– SNPs in Coding (Amino Acid Changes)
– Linkage disequilibrium (SNPs in other functional regions, i.e.
regulatory elements)
cSNPs - Both Deep and Average Coverage
Available from the PGAs
CD36 - Southwestern PGA - Deep cSNP Discovery Strategy Healthy, High Cholesterol, High Triglycerides, Congential Cardiac
Abnormalities, Left Ventricular Hypertrophy …….
CD36 - SeattleSNPs PGA - Average cSNP Discovery Strategy Healthy only
SIFT (Sorting Intolerant From Tolerant) Coding Changes
CYP4F2
Trp (W)  Gly (G)
Predicted to be tolerated
Val (V)  Gly (G)
Predicted not to be tolerated
Ng and Henikoff, Gen. Res. 2002
SNP-Based Association Studies
Indirect: Use dense map of SNPs and test for linkage
disequilibrium (use association to find sites in
entire sequence (non-coding) with function)
5’
Arg-Cys
Val-Val
Collins, Guyer, Chakravarti Science 278:1580-81, 1997
3’
SNP Discovery and Genotyping Workshop
• SNP discovery strategies
Debbie Nickerson
• Identifying SNPs by association for genotypephenotype analysis of candidate genes
Chris Carlson
• Identifying haplotypes for genotype-phenotype
analysis of candidate genes
Dana Crawford
• SNP genotyping strategies
Debbie Nickerson
Selecting SNPs for
Genotype-Phenotype Analysis
Using Allelic Association
(Linkage Disequilibrium)
Christopher Carlson
[email protected]
Candidate Gene Association
Analysis
• Describe existing genetic variation
– Rare SNPs (deep exonic resequencing)
– Common SNPs (complete resequencing)
• Select a subset of SNPs for genotyping
– cSNPs (amino acid changes)
– htSNPs (resolve haplotypes)
– tagSNPs (patterns of genotype)
• Test for genotype/phenotype correlations
SeattleSNPs Resequencing Strategy I
• Resequence the complete genomic region of each gene
– 2000 bp upstream of first exon
– 1500 bp downstream of poly-A signal
– All exons and introns for genes below 35 kbp
Image courtesy of GeneSNPs
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
SeattleSNPs Resequencing Strategy II
• Resequence candidate genes from
inflammation and coagulation
pathways
• Resequence 47 individuals
– 24 African American
– 23 European American
Homozygote common
Heterozygote
Homozygote rare
Missing Data
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
VG2
• Visual Genotype 2
–
–
–
–
Web interface
Visualize genotypes
View SNPs by frequency
Sort on similarity
between sites
– Sort on similarity
between samples
– Visualize LD
Preliminary Analyses
• Hardy Weinberg
Equilibrium
• Population specificity
• Nucleotide diversity
• Pop genetics statistics
(e.g. Tajima’s D)
SNP Selection: cSNPs
• Genotype SNPs which change amino acids
• Genotype other “good story” SNPs
– SNPs in known regulatory elements
– SNPs in Conserved Noncoding Sequences
Image courtesy of GeneSNPs
SNP Selection: htSNPs
• Genotype “haplotype
tagging” SNPs which
resolve existing
common haplotypes
SNP Selection: htSNPs
• Genotype “haplotype
tagging” SNPs which
resolve existing
common haplotypes
SNP Selection: tagSNPs
• Resequence a modest number of samples
– Describe patterns of genotype at all
common SNPs
– Genotype tagSNPs which efficiently
capture existing patterns of genotype
Linkage Disequilibrium
A B
Haplotype is the pattern of alleles
on a single chromosome
– 4 possible haplotypes
Linkage Disequilibrium (LD)
describes the allelic association
between two SNPs
Two popular LD statistics:
D´
r2
Complete LD
A B
Unequal allele frequency
Allelic association is as strong as
possible
– 3 haplotypes observed
– No detected recombination
between SNPs
– Genotype is not perfectly
correlated
D´ = 1
r2 < 1
Perfect LD
A B
Equal allele frequency
Allelic association is as strong as
possible
– 2 haplotypes observed
– No detected recombination
between SNPs
– Genotype is perfectly correlated
D´ = 1
r2 = 1
Rational SNP Selection
Select SNPs to genotype on the basis of LD
• Some SNPs are in LD with many other SNPs
• Some SNPs are in LD with no other SNPs
• SNPs between a pair of associated SNPs are not
necessarily associated with the flanking SNPs
LD SNP Selection Example
CSF3 in European Americans
•5200 bp
•17 SNPs
LD SNP Selection Example
CSF3 in European Americans
•5200 bp
•17 SNPs
•10 common SNPs (above
10% minor allele
frequency)
LD Site Selection Algorithm
• Find minimal set of SNPs
for assay, such that
each SNP is either
assayed directly or
above r2 threshold with
an assayed SNP
•Calculate all pairwise r2
values
•Set r2 threshold based on
power estimates for
study
LD Site Selection Algorithm
• Find minimal set of SNPs
for assay, such that
each SNP is either
assayed directly or
above r2 threshold with
an assayed SNP
•Calculate all pairwise r2
values
•Set r2 threshold based on
power estimates for
study
CSF3 Site Selection
• Threshold LD: r2 > 0.64
– Bin 1: 4 sites
– Bin 2: 4 sites
– Bin 3: 2 sites
• Genotype 1 SNP from
each bin, chosen for
biological intuition or
ease of assay design
Power and LD
• Given
– All common SNPs described
– Patterns of LD between common SNPs are known
• Select SNPs such that every SNP is either
– Directly assayed
– Associated with an assayed SNP
• Test for disease associations with assayed SNPs
• Power to detect disease associations at unassayed
SNPs depends on r2 between assayed and
unassayed SNPs
LD Selection and Haplotype
• LD selected SNPs provide the highest possible
haplotype diversity for a given number of SNPs
assayed
• LD selection is robust to recombination and
hotspot structure
• LD selection is sensitive to population
stratification
SNP Selection Summary
• It is possible to test all common variants in
a candidate gene directly for risk
association (main effects) with meaningful
null negative results
• Caveat: Higher order risks unaddressed
– Haplotype (G X G effects within a locus)
– Epistasis (G X G effects between loci)
– Environment (G X E effects)
SNP Discovery and Genotyping Workshop
• SNP discovery strategies
Debbie Nickerson
• Identifying SNPs by association for genotypephenotype analysis of candidate genes
Chris Carlson
• Identifying haplotypes for genotype-phenotype
analysis of candidate genes
Dana Crawford
• SNP genotyping strategies
Debbie Nickerson
Identifying Haplotypes for
Genotype-Phenotype
Analysis
Dana C. Crawford
[email protected]
Outline of discussion
• Constructing or inferring haplotypes
• Haplotype tools available in PGA
• Description of haplotypes in SeattleSNPs genes
• Use of VH1 tool to visually inspect
– Haplotype blocks
– Haplotype diversity
– Hotspots of recombination
• Summary of SeattleSNPs haplotype data
What is a Diplotype ?
• Humans are diploid
• At each SNP there are two alleles,
which are observed as a genotype
• At each gene there are two
haplotypes, which are observed as
a multi-site genotype, or diplotype
What is a Haplotype?
A: “…a unique combination of genetic markers present
in a chromosome.” pg 57 in Hartl & Clark, 1997
VH1 – haplotype visualization tool
How Do You Construct Haplotypes?
1. Collect extended family members
C/C, A/G
TT
GG
T/T, G/G
C/T, A/A
C/C, A/G
C/T, A/G
CT
AG
CC
AG
How Do You Construct Haplotypes?
2. Go from diploid
to haploid via
somatic cell hybrids
e.g. Patil et al 2001
How Do You Construct Haplotypes?
3. Allele-specific PCR
SNP 1
C/T
SNP 2
A/G
How Do You Construct Haplotypes?
4. Statistical inference
•
Clark Algorithm
•
EM (Arlequin)
•
Phase Ligation (HAPLOTYPER)
•
PHASE
Clark Algorithm
• Find unambiguous
haplotypes
– Homozygotes
– Single Heterozygotes
Clark Algorithm
• Find ambiguous diplotypes
formed from two
unambiguous genotypes
Clark Algorithm
• Find ambiguous diplotypes formed
from one unambiguous genotype
and one new genotype
Clark Algorithm
• Iterate until either all haplotypes
resolve, or ambiguous
haplotypes are inconsistent with
any inferred haplotype
Haplotype Algorithm
Comparison
• Clark
– Intuitive
– Fast
• PHASE
– Complete solution
– Slightly more
accurate than EM
– Slow version 2 faster
• EM
– Complete solution
– Slightly more
accurate than Clark
– Robust to ambiguity
• Haplotyper (Ligation)
– Fast
– Better than Clark
– Less accurate than
EM or PHASE
Haplotype Tools in the PGA
InnateImmunity
• 25 genes re-sequenced in innate immunity pathway
• 4 populations: European and African-Americans,
Hispanics, Asthmatics
• PHASE and Haplotyper results posted on website
http://innateimmunity.net
Haplotype Tools in the PGA
SeattleSNPs
•
•
•
•
120 genes re-sequenced in inflammation response
2 populations: European- and African-Americans
PHASE results posted on website
Interactive tool (VH1) to visualize and sort haplotypes
http://pga.gs.washington.edu
Distribution of Haplotypes in
100 SeattleSNPs Genes
50
45
40
Number of haplotypes
35
30
25
AD
20
15
ED
10
5
0
0
10
20
30
40
50
60
Number of genes
70
80
90
100
Common Haplotypes in 100
SeattleSNPs Genes
(Frequency >5%)
Population
>5% MAF
Average
Range
ED
4.54
1-8
AD
4.99
0 - 11
Haplotype Sharing Between
Populations in 100 SeattleSNPs Genes
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Non-shared
Shared
ED
AD
Number of Haplotypes From Two
Different Discovery Strategies
Average number of haplotypes per gene
35
30
25
All SNPs
>5%
20
15
Coding
SNPs,
>5%
10
5
0
AD
ED
Combined
Haplotype Structures Are Similar
Across Discovery Strategies…
FGB – African-Americans
Coding SNPs
29 SNPs >5%
13 SNPs >5%
…But, Not For All Genes
F10 – African-Americans
Coding SNPs
48 SNPs >5%
13 SNPs >5%
Are Blocks Preserved Using
Different Discovery Strategies?
Four-gamete test:
A
B
a
b
A
B
a
A
b
b
a
B
Yes*, for some:
10% of genes in AD
25% of genes in ED
Fewer “blocks” with fewer SNPs/kb
HaploBlockFinder; Zhang and Jin 2003
*>75%
of the blocks are preserved
Using Visualization Tools (VH1)
To Identify Haplotype Blocks
IL10:
• Rare sites removed
• Sorted by related sites
• “Block” structure evident
Using VH1 to Identify Highly
Divergent Haplotypes
• Some haplotypes are highly divergent
• More likely to have functional consequences?
• Mixed Blessing:
– Easier to detect
– Harder to dissect
Using
Haplotypes To
Identify
Hotspots Of
Recombination
CD36 haplotypes,
sorted by sample
Linkage Disequilibrium and Hotspots
Associated Sites
CD36
Hotspot in
between
sites need to be
typed from both
ends
Detection of Recombination Hotspots
In Candidate Genes
HOTSPOTTER
• Developed by Na Li and Matthew Stephens
• Multilocus model for LD:
Does not rely on “block-like” patterns
Relates LD to underlying recombination process
Incorporated into new version of PHASE (v2.0)
students.washington.edu/lina/software/
CD36 – combined population
CD36 – AD and ED populations
HOTSPOTTER
Preliminary Results
15 out of 100 genes have evidence of a hotspot:
AGTR1
APOB
CD36
IL1B
IL21R
IL4
NOS3
PLAUR
PON1
SERPIN45
SELP
SFPA2
SFTPB
VCAM1
VEGF
SeattleSNPs Haplotype Summary
• More haplotypes per gene than previously described
• <50% of African-American chromosomes are represented
by common shared haplotypes
• Block structure is preserved across discovery strategies
for only a fraction of the genes
• Evidence for hotspots of recombination in human genes
SNP Discovery and Genotyping Workshop
• SNP discovery strategies
Debbie Nickerson
• Identifying SNPs by association for genotypephenotype analysis of candidate genes
Chris Carlson
• Identifying haplotypes for genotype-phenotype
analysis of candidate genes
Dana Crawford
• SNP genotyping strategies
Debbie Nickerson
Ideals for SNP Genotyping
• High Sensitivity - PCR but moving towards direct
genomic DNA detection
• High Specificity - Accurate
• Simple process - Easy to automate - High Throughput
• Multiplexing - Perform many assays at once - decrease costs
• Cheap
SNP Genotyping
Matched
Probe and Target
C Allele
C
Target
C
G
Mis-Matched
T Allele
C
Allele-Specific Hybridization
Hybridize
+ddCTP
Polymerase Extension
Target
Oligonucleotide Ligation
C
Target
Invader
Taqman
C
Target
C
Target
A
Fail to hybridize
C
G
A
C incorporat ed
C Fails t o incorporate
C
G
A
Ligate
Fail to ligat e
C
G
A
Cleave
Fail to cleave
C
G
Degrade
C
A
Fail t o degrade
C
Allele-Specific PCR
C
Target
G
A
Amplify
Fail to amplify
SNPTyping Formats
Microtiter Plates - Fluorescence
eg. Taqman - Good for a few markers - lots of
samples - PCR
Size Analysis by Mass or Electrophoresis
eg. Sequenom or SnapShot - Moderate Multiplexing
reducing costs
Arrays - Custom or Universal
eg. Affymetrics, Illumina or ParAllele - Highly multiplexed HighThroughput - Genotype directly on
genomic DNA
Taqman
Genotyping with fluorescence-based homogenous assays
(single-tube assay)
A
G
Genotype Calling - Cluster Analysis
Genotyping by Mass Spectrometry
Multiplex ~ 5 SNPs
Comparative Genotyping in Populations
Pooled DNA
Population 1
PCR Pooled DNA
Quantitative Assay
Estimate Allele Frequency
Polymorphism
60/40
Pooled DNA
Population 2
PCR Pooled DNA
Quantitative Assay
Estimate Allele Frequency
Polymorphism
85/15
Pooled Genotyping
Advantages:
Speed, Cost
Major Disadvantages:
Loss of haplotype information
Loss of stratification by phenotype
or environmental factors
SNP Genotyping
Custom SNP Genotyping Chips:
Multiplexed
C
Genotyping
- Universal Tag Readouts
G
A
T
Locus 2 Specific Sequence
Locus 1 Specific Sequence
Tag1 sequence
Tag2 sequence
cTag1 sequence
cTag2 sequence
Subst rate
Bead or Chip
Substrate
Bead or Chip
Bead Array
Chip Array
Tag 1
Tag 2
Tag 3
Tag 4
Illumina
Multiplex ~1,000 SNPs
Not dependent on primary PCR
ParAllele
Illumina Genotyping - Gap Ligation
1,000 SNPs Assayed on 96 Samples
SNP Genotyping
Lots of systems - Still costly but dropping
Offering Moderate to High throughputs
Systems vary in price $$ -$$$$
Laboratory Information Management Systems
(Key: Track - Samples,
- Assays
- Completion rate
- Reproducibility/Error Analysis)