Cryptic Variation in the Human mutation rate

1
Cryptic Variation in the Human mutation rate
Alan Hodgkinson Adam Eyre-Walker, Manolis
Ladoukakis
2
Variation in the mutation rate

• Between different chromosomes
• Between regions on chromosomes
• Neighbouring nucleotides

3
Simple context effects
Hwang and Green (2004) PNAS 101 13994-14001
4
Cryptic Variation

• Remote context
• AGTCGGTTACCGTGACGTTGAACGTGT

5
Cryptic Variation

• Remote context
• AGTCGGTTACCGTGACGTTGAACGTGT
• Degenerate context
• AGTCGGTTACCGTGYSRGYGAACGTGT

6
Cryptic Variation

• Remote context
• AGTCGGTTACCGTGACGTTGAACGTGT
• Degenerate context
• AGTCGGTTACCGTGYSRGYGAACGTGT
• No context / Complex context

7
Our approach to the problem

• Search for SNPs in human sequences that also
  have a SNP in the orthologous position in chimp.

8
Our approach to the problem

• Search for SNPs in human sequences that also
  have a SNP in the orthologous position in chimp.

Do we see more coincident SNPs than expected by
chance?
9
The method

• Extract all human SNPs from dbSNP and construct
  a BLAST database on a chromosome by chromosome
  basis.

10
The method

• Extract all human SNPs from dbSNP and construct
  a BLAST database on a chromosome by chromosome
  basis.
• Extract all chimp SNPs from dbSNP with 50bp
  either side of SNP.

11
The method

• Extract all human SNPs from dbSNP and construct
  a BLAST database on a chromosome by chromosome
  basis.
• Extract all chimp SNPs from dbSNP with 50bp
  either side of SNP.
• BLAST chimp SNPs against human database.

12
The method

• Extract all human SNPs from dbSNP and construct
  a BLAST database on a chromosome by chromosome
  basis.
• Extract all chimp SNPs from dbSNP with 50bp
  either side of SNP.
• BLAST chimp SNPs against human database.
• Extract results above a certain level of
  homology where there is a SNP on both sequences
  and reduce to 40bp either side of central
  position.

13
The method

• Extract all human SNPs from dbSNP and construct
  a BLAST database on a chromosome by chromosome
  basis.
• Extract all chimp SNPs from dbSNP with 50bp
  either side of SNP.
• BLAST chimp SNPs against human database.
• Extract results above a certain level of
  homology where there is a SNP on both sequences
  and reduce to 40bp either side of central
  position.
• Repeating both including and excluding CpG
  effects.

14
Results

• 1.5 million chimp SNPs.
• 310,000 81bp alignments containing a human and
  chimp SNP.

15
Results

• 1.5 million chimp SNPs.
• 310,000 81bp alignments containing a human and
  chimp SNP.
• Observe the number of coincident SNPs.
• Calculate the expected number, taking into
  account the effects of neighbouring nucleotides.

16
Results
Obs Exp Ratio
All 11571 6592 1.76 (1.72,1.79)
No-CpG 5028 2533 1.98 (1.93,2.04)
17
Results
C/T G/A C/A G/T C/G A/T
C/T 1.91 1.04 1.19 1.21 0.96
G/A 1.83 1.24 1.02 1.14 1.40
C/A 1.23 1.08 4.81 1.28 1.39
G/T 1.15 1.38 4.95 1.27 0.77
C/G 1.09 1.14 1.24 1.40 2.79
A/T 0.94 1.06 1.79 0.99 15.43
18
Alternative Explanations

• Bias in the Method
• Selection
• Ancestral Polymorphism
• Paralogous SNPs

19
Alternative Explanations

• Bias in the Method
• Selection
• Ancestral Polymorphism
• Paralogous SNPs

20
Methodological Bias

• Simulated data with same density of human and
  chimp SNPs as dbSNP under different divergence
  and mutation patterns.
• Method worked well under realistic conditions.

21
Methodological Bias
All sites (HG)
Div Obs Exp Ratio 95 CI
0 839 812 1.033 (0.963,1.103)
1 2419 2316 1.040 (1.003,1.086)
2 681 685 0.995 (0.920,1.069)
Non CpG sites (HG)
Div Obs Exp Ratio 95 CI
0 401 428 0.936 (0.844,1.028)
1 1182 1228 0.963 (0.908,1.018)
2 374 400 0.935 (0.840,1.030)
22
Methodological Bias
All sites (HG)
Div Obs Exp Ratio 95 CI
0 839 812 1.033 (0.963,1.103)
1 2419 2316 1.040 (1.003,1.086)
2 681 685 0.995 (0.920,1.069)
Non CpG sites (HG)
Div Obs Exp Ratio 95 CI
0 401 428 0.936 (0.844,1.028)
1 1182 1228 0.963 (0.908,1.018)
2 374 400 0.935 (0.840,1.030)
23
Alternative Explanations

• Bias in the method
• Selection
• Ancestral Polymorphism
• Paralogous SNPs

24
Selection

• Areas of low SNP density result in clustering

Human
Chimp
25
Selection

• Areas of low SNP density result in clustering

Human
Chimp
Apparent excess of coincident SNPs
26
Selection

• No clustering

27
Alternative Explanations

• Bias in the method
• Selection
• Ancestral Polymorphism
• Paralogous SNPs

28
Ancestral Polymorphism

• SNP inherited from common ancestor of chimp and
  human

29
Ancestral Polymorphism

• SNP inherited from common ancestor of chimp and
  human

Increase in coincident SNPs
30
Ancestral Polymorphism

• Expect observed/expected ratio to be same for
  all transitions

C/T G/A C/A G/T C/G A/T
C/T 1.91 1.04 1.19 1.21 0.96
G/A 1.83 1.24 1.02 1.14 1.40
C/A 1.23 1.08 4.81 1.28 1.39
G/T 1.15 1.38 4.95 1.27 0.77
C/G 1.09 1.14 1.24 1.40 2.79
A/T 0.94 1.06 1.79 0.99 15.43
31
Ancestral Polymorphism

• Repeated initial analysis with macaque data.
• Humans and Macaque split 23-24 million years
  ago so we expect there to be no shared
  polymorphisms.

32
Ancestral Polymorphism

• Repeated initial analysis with macaque data.
• Humans and Macaque split 23-24 million years
  ago so we expect there to be no shared
  polymorphisms.

Obs Exp Ratio
All 77 47 1.64 (1.27,2.00)
No-CpG 34 23 1.51 (1.001,2.02)
33
Alternative Explanations

• Bias in the method
• Selection
• Ancestral Polymorphism
• Paralogous SNPs

34
Paralogous SNPs

• Excess of coincident SNPs a consequence of
  artifactual SNPs called as a result of
  substitutions in paralogous regions.

35
Paralogous SNPs

• Excess of coincident SNPs a consequence of
  artifactual SNPs called as a result of
  substitutions in paralogous regions.
• Musumeci et al (2010) 8.32 of human variation
  in dbSNP may be due to paralogy.

36
Paralogous SNPs

• Excess of coincident SNPs a consequence of
  artifactual SNPs called as a result of
  substitutions in paralogous regions.
• Musumeci et al (2010) 8.32 of human variation
  in dbSNP may be due to paralogy.

AGCTGCACGT Y CGGCATCCAA SNP AGCTGCACGT T
CGGCATCCAA Chromosome 1 AGCTGCACGT A
CGGCATCCAA Chromosome 7
Artifactual SNP
37
Paralogous SNPs
AGCTGCACGT (T/A) CGGCATCCAA AGCTGCACGT T
CGGCATCCAA
AGCTGCACGT (T/A) CGGCATCCAA AGCTGCACGT T
CGGCATCCAA AGCTGCACGT A CGGCATCCAA
38
Paralogous SNPs
AGCTGCACGT (T/A) CGGCATCCAA AGCTGCACGT T
CGGCATCCAA
AGCTGCACGT (T/A) CGGCATCCAA AGCTGCACGT T
CGGCATCCAA AGCTGCACGT A CGGCATCCAA
3.6 of coincident SNPs are possibly a
consequence of paralogous sequences
39
Alternative Explanations

• Bias in the method
• Selection
• Ancestral Polymorphism
• Paralogous SNPs

Cryptic variation in the mutation rate
40
Context Analysis

• 4517 sequences containing non-CpG coincident
  SNPs flanked by 200bp.
• Tabulate triplet frequencies at each position in
  surrounding sequences.
• Test whether the proportions of triplets we
  observe at each position significantly different
  from the proportions in the sequences as a whole.

41
Context Analysis

• Coincident SNP in central position

42
Context Analysis

• Coincident SNP in central position

No obvious context surrounding coincident SNPs
43
Genomic Distribution

• Tallied the number of coincident SNPs per MB
• 3.91 coincident SNPs per MB.
• 1.68 non-CpG coincident SNPs per MB.

44
Genomic Distribution

• Tallied the number of coincident SNPs per MB
• 3.91 coincident SNPs per MB.
• 1.68 non-CpG coincident SNPs per MB.

• If randomly distributed expect Poisson
  distribution and ? ?2 3.91

45
Genomic Distribution

• Tallied the number of coincident SNPs per MB
• 3.91 coincident SNPs per MB.
• 1.68 non-CpG coincident SNPs per MB.

• If randomly distributed expect Poisson
  distribution and ? ?2 3.91
• ?2 13.27 (plt0.001) and so sampling variance
  explains approximately 30 of total variance.

46
Genomic Distribution
Feature r r2 p
SNP density 0.256 0.0655 lt0.001
Distance to Telomere -0.022 0.0004 0.226
Distance to Centromere 0.011 0.0001 0.565
Recombination Rate 0.107 0.0114 lt0.001
Nucleosome Association 0.004 0.0000 0.832
Gene Density -0.022 0.0004 0.230
GC content -0.006 0.0000 0.741
47
Genomic Distribution

• SNP densities must drive coincident SNP
  densities to a certain extent as approximately
  half of coincident SNPs are created by chance
  alone.

48
Genomic Distribution

• SNP densities must drive coincident SNP
  densities to a certain extent as approximately
  half of coincident SNPs are created by chance
  alone.
• Recombination rate positively correlated with
  SNP density (r 0.242, plt0.001).
• Partial correlation controlling for SNP density
  r 0.048, p0.011.

49
Genomic Distribution

• SNP densities must drive coincident SNP
  densities to a certain extent as approximately
  half of coincident SNPs are created by chance
  alone.
• Recombination rate positively correlated with
  SNP density (r 0.242, plt0.001).
• Partial correlation controlling for SNP density
  r 0.048, p0.011.
• SNP densities explain 6.5 of the variance,
  recombination rate explains 0.2 of the variance
  of coincident SNPs.

50
Genomic Distribution
Feature r r2 p
Coincident SNP Density 0.256 0.0655 lt0.001
Distance to Telomere -0.171 0.0292 lt0.001
Distance to Centromere -0.047 0.0022 0.012
Recombination Rate 0.234 0.0548 lt0.001
Nucleosome Association 0.187 0.0350 lt0.001
Gene Density 0.064 0.0041 0.001
GC content 0.184 0.0339 lt0.001
51
Quantification

• Use Log-normal distribution of relative mutation
  rates due to cryptic variation.
• Model the number of coincident SNPs under the
  effects of cryptic variation.
• Incorporate effects of divergence.

52
Quantification

• Use Log-normal distribution of relative mutation
  rates due to cryptic variation.
• Model the number of coincident SNPs under the
  effects of cryptic variation.
• Incorporate effects of divergence.

What level of variation in the log-normal
distribution explains our results?
53
Log-normal model
Fastest 5 of sites mutate 16.4 times faster
than slowest 5 of sites.
54
Summary

• Cryptic variation in the mutation rate.

55
Summary

• Cryptic variation in the mutation rate.
• No obvious context surrounding coincident SNPs.

56
Summary

• Cryptic variation in the mutation rate.
• No obvious context surrounding coincident SNPs.
• Variation is truly cryptic.

57
Summary

• Cryptic variation in the mutation rate.
• No obvious context surrounding coincident SNPs.
• Variation is truly cryptic.
• Genomic distribution of coincident SNPs is
  over-dispersed

58
Summary

• Cryptic variation in the mutation rate.
• No obvious context surrounding coincident SNPs.
• Variation is truly cryptic.
• Genomic distribution of coincident SNPs is
  over-dispersed
• Variation in mutation rate is substantial.

59
Acknowledgments

• BBSRC
• People

Manolis Ladoukakis
Adam Eyre-Walker