Making sense of DNA methylation data

PhD completion seminar presented at the Walter Eliza Hall Institute on 15 September, 2014.
Making sense of DNA methylation data by Peter Hickey is licensed under a Creative Commons Attribution 4.0 International License.

Transcript

1. Making  sense  of     DNA  methyla4on  data   Peter  Hickey     @PeteHaitch   15  September  2014  
2. “Oh,  you’re  a  sta/s/cian…”   How  I  spend  my  4me  
3. Coding RA work Writing Analyses Talk preparation Seminars Exercise Japanese study Reading papers Email and admin Tutoring Meetings 0 200 400 600 Hours How I've spent my time since 13 January, 2014
4. How  I  spend  my  4me   hEp://gifurl.com/gifs/1145  
5. How  I  spend  my  4me   hEp://wifflegif.com/gifs/511132-­‐richard-­‐ayoade-­‐the-­‐it-­‐crowd-­‐gif  
6. Project  aim   How  to  analyse  whole-­‐genome  bisulfite-­‐sequencing   experiments  to  learn  about  DNA  methyla<on?    
7. hEps://www.flickr.com/photos/theloushe/4640871734   Exploratory  data  analyses  
8. hEp://tstoaddicts.files.wordpress.com/2013/08/simpsons-­‐ques4ons.gif   Developing  methods  
9. hEp://www.mhjaso.com/blog/wp-­‐content/uploads/2014/06/Homer-­‐Tries-­‐to-­‐Build-­‐a-­‐Barbeque-­‐Set.jpg   Implemen4ng  methods  
10. hEp://cdn.amazinganimalstories.com/wp-­‐content/uploads/2013/11/ATT00038.jpg   Collabora4on  
11. 5mCs  and  one  assay  
12. ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC!
13. ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC!
14. m m m m .! ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC! m m m.!
15. "Cytosine  becomes  thymine"  by  CFCF  -­‐  Own  work.  Licensed  under  Crea4ve  Commons  AEribu4on-­‐Share  Alike  3.0  via  Wikimedia  Commons  -­‐   hEp://commons.wikimedia.org/wiki/File:Cytosine_becomes_thymine.png#mediaviewer/File:Cytosine_becomes_thymine.png  
16. "Cytosine  becomes  thymine"  by  CFCF  -­‐  Own  work.  Licensed  under  Crea4ve  Commons  AEribu4on-­‐Share  Alike  3.0  via  Wikimedia  Commons  -­‐   hEp://commons.wikimedia.org/wiki/File:Cytosine_becomes_thymine.png#mediaviewer/File:Cytosine_becomes_thymine.png  
17. ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC!
18. m m m m .! ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC! m m m.!
19. hEp://episona.com/wp-­‐content/uploads/2014/01/Agou4-­‐Mouse-­‐1024x802.jpg  
20. hEp://www.ous-­‐research.no/kristensen/images/projects/Promoter_methyla4on.jpg   Expressed   Not  expressed  
21. hEp://www.ous-­‐research.no/kristensen/images/projects/Promoter_methyla4on.jpg   Expressed   Not  expressed  
22. Cancer  Genome  Atlas  Research  Network.  "Integrated  genomic  analyses  of   ovarian  carcinoma."  Nature  474.7353  (2011):  609-­‐615.   AMT   Gene  expression   Promoter  methyla4on   0   1   -­‐4   0   2   -­‐2  
23. Cancer  Genome  Atlas  Research  Network.  "Integrated  genomic  analyses  of   ovarian  carcinoma."  Nature  474.7353  (2011):  609-­‐615.   AMT   Gene  expression   Promoter  methyla4on   0   1   -­‐4   0   2   -­‐2   hEp://commons.wikimedia.org/wiki/File:Calico_cat_-­‐ _Phoebe.jpg#mediaviewer/File:Calico_cat_-­‐_Phoebe.jpg  
24. Cancer  Genome  Atlas  Research  Network.  "Integrated  genomic  analyses  of   ovarian  carcinoma."  Nature  474.7353  (2011):  609-­‐615.   AMT   Gene  expression   Promoter  methyla4on   0   1   -­‐4   0   2   -­‐2   hEp://commons.wikimedia.org/wiki/File:Calico_cat_-­‐ _Phoebe.jpg#mediaviewer/File:Calico_cat_-­‐_Phoebe.jpg   hEp://commons.wikimedia.org/wiki/File%3ABlastocyst_embryo.png  
25. m m m m .! ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC! m m m! ! ACGTGAAACGAACTATCGG! m m m.!
26. m m m m .! ACGCGAAACGTTCTATCGG! TGCGCTTTGCAAGATAGCC! m m m! ! ACGTGAAACGAACTATCGG! m m m.!
27. m m m m .! ACGCGAAACGTTCTATCGG! + ! Sodium  bisulfite   =   ACGTGAAACGAACTATCGG! m m m.!
28. m m m m .! ACGCGAAACGTTCTATCGG! + ! PCR  amplifica4on   =   ACGTGAAACGAACTATCGG! m m m.!
29. m m m m .! ACGCGAAACGTTCTATCGG! + ! PCR  amplifica4on   =   ACGCGAAACGTTCTATCGG! m m m.!
30. m m m m .! ACGCGAAACGTTCTATCGG! + ! Sodium  bisulfite   =   ACGTGAAACGAACTATCGG! m m m.!
31. m m m m .! ACGCGAAACGTTCTATCGG! + ! Sodium  bisulfite   =   ACGTGAAACGAACTATCGG! m m m.!
32. m m m m .! ACGCGAAACGTTCTATCGG! + ! Sodium  bisulfite   =   ACGUGAAACGTTCTATCGG! m m m.!
33. m m m m .! ACGUGAAACGTTCTATCGG! + ! PCR   =   ACGTGAAACGAACTATCGG! m m m.!
34. m m m m .! ACGUGAAACGTTCTATCGG! + ! PCR  amplifica4on   =   ACGTGAAACGAACTATCGG! m m m.!
35. m m m m .! ACGUGAAACGTTCTATCGG! + ! PCR  amplifica4on   =   ACGTGAAACGTTCTATCGG! m m m.!
36. m m m m .! ACGUGAAACGTTCTATCGG! + ! PCR  amplifica4on   =   ACGTGAAACGTTCTATCGG! m m m.!
37. Bisulfite  treatment  of  DNA   +           =   Whole-­‐genome  bisulfite-­‐sequencing  
38. ACGCGAAACGTTCTATCG
39. ACGCGAAACGTTCTATCG
44. β 1 = 3/3
45. β 2 = 4/4
46. β 3 = 2/4
47. β 4 = 0/4
48. 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 25 50 75 sample sample1 sample2 sample3 sample4 sample5
49. CpG island 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 25 50 75 sample sample1 sample2 sample3 sample4 sample5
50. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 25 50 75 sample sample1 sample2 sample3 sample4 sample5
51. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Coverage 0 20 40 60 sample sample1 Data:  Emma  Whitelaw  
52. Lister  data   ADS   ADS-­‐adipose   ADS-­‐iPSC   Organism   Human  (female)   Human  (female)   Human  (female)   Cell  type   Soma4c   Soma4c   Induced  pluripotent  stem   cell  (iPSC)   Descrip/on   Adipose   Adipocytes  derived  from   ADS   iPSC  line  derived  from  ADS   Sequencing   75  bp  paired-­‐end   75  bp  paired-­‐end   75  bp  paired-­‐end   Average   coverage   23×   24×   26×     Lister,  Ryan,  et  al.  "Hotspots  of  aberrant  epigenomic  reprogramming  in  human  induced  pluripotent  stem   cells."  Nature  471.7336  (2011):  68-­‐73.  
53. ADS ADS−adipose ADS−iPSC 0 2,000,000 4,000,000 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 β count
54. ADS ADS−adipose ADS−iPSC 0 2,000,000 4,000,000 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 β count CpG island Non CpG island
55. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Coverage 0 20 40 60 sample sample1 Data:  Emma  Whitelaw  
56. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Coverage 0 20 40 60 sample sample1 Data:  Emma  Whitelaw  
57. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Coverage 0 20 40 60 sample sample1 Data:  Emma  Whitelaw  
58. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 20 40 60 sample mouse1 Data:  Emma  Whitelaw  
59. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 25 50 75 sample mouse1 mouse2 mouse3 mouse4 mouse5 Data:  Emma  Whitelaw  
60. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 25 50 75 sample mouse1 mouse2 mouse3 mouse4 mouse5 Data:  Emma  Whitelaw  
61. Co-­‐methyla/on  
62. Co-­‐methyla4on  =  co-­‐occurence   “The  presence  of  methyla<on  over  a  stretch  of   neighboring  CpG  posi<ons”   Schatz,  Philipp,  Dimo  Dietrich,  and  MaEhias  Schuster.  "Rapid  analysis  of  CpG  methyla4on  paEerns  using  RNase  T1  cleavage  and  MALDI-­‐TOF."   Nucleic  acids  research  32.21  (2004):  e167-­‐e167.  
63. Co-­‐methyla4on  =  co-­‐occurence   “The  presence  of  methyla<on  over  a  stretch  of   neighboring  CpG  posi<ons”   Schatz,  Philipp,  Dimo  Dietrich,  and  MaEhias  Schuster.  "Rapid  analysis  of  CpG  methyla4on  paEerns  using  RNase  T1  cleavage  and  MALDI-­‐TOF."   Nucleic  acids  research  32.21  (2004):  e167-­‐e167.  
64. Co-­‐methyla4on  =  co-­‐occurence   “The  presence  of  methyla<on  over  a  stretch  of   neighboring  CpG  posi<ons”   Schatz,  Philipp,  Dimo  Dietrich,  and  MaEhias  Schuster.  "Rapid  analysis  of  CpG  methyla4on  paEerns  using  RNase  T1  cleavage  and  MALDI-­‐TOF."   Nucleic  acids  research  32.21  (2004):  e167-­‐e167.  
65. Co-­‐methyla4on  =  correla4on   “The  rela<onship  between  the  degree  of   methyla<on  over  distance”   1.  Within-­‐fragment  co-­‐methyla4on   2.  Correla4on  of  β-­‐values   Eckhardt,  Florian,  et  al.  "DNA  methyla4on  profiling  of  human  chromosomes  6,  20  and  22."  Nature  gene4cs  38.12  (2006):  1378-­‐1385.  
66. Co-­‐methyla4on  =  correla4on   “The  rela<onship  between  the  degree  of   methyla<on  over  distance”   1.  Within-­‐fragment  co-­‐methyla4on   2.  Correla4on  of  β-­‐values   Eckhardt,  Florian,  et  al.  "DNA  methyla4on  profiling  of  human  chromosomes  6,  20  and  22."  Nature  gene4cs  38.12  (2006):  1378-­‐1385.  
67. Co-­‐methyla4on  =  correla4on   “The  rela<onship  between  the  degree  of   methyla<on  over  distance”   1.  Within-­‐fragment  co-­‐methyla4on   2.  Correla4on  of  β-­‐values   Eckhardt,  Florian,  et  al.  "DNA  methyla4on  profiling  of  human  chromosomes  6,  20  and  22."  Nature  gene4cs  38.12  (2006):  1378-­‐1385.  
68. Co-­‐methyla4on  =  correla4on   “The  rela<onship  between  the  degree  of   methyla<on  over  distance”   1.  Within-­‐fragment  co-­‐methyla/on   2.  Correla4on  of  β-­‐values   Eckhardt,  Florian,  et  al.  "DNA  methyla4on  profiling  of  human  chromosomes  6,  20  and  22."  Nature  gene4cs  38.12  (2006):  1378-­‐1385.  
69. DATA TYPE 38,721,970 bp 38,721,980 bp 38,721,990 bp 38,722,000 bp 38,722,010 bp 38,722,020 bp 38,722,030 bp 38,722,040 bp 38,722,050 bp 91 bp chr4 p16.2 p15.33 p15.31 p15.1 p14 p13 p11 q12 q13.1 q13.2 q21.1 q21.23 q22.2 q23 q24 q25 q26 q27 q28.2 q31.1 q31.22 q31.3 q32.2 q33 q34.2 q35.1 IMR90-iPSC coverage IMR90-iPSC [0 - 35] G A Sequence RefSeq genes CpGs CpG islands A G A G T A G C A A A C A C T A A T C T A C G A A T A A T G A A C A T A G G C A T T A T T T T A A G A A C C A A A A G A A A G C A C G T G G G C A T T T G G T T T A C A C A T C A C T CpGs  
70. NAME DATA TYPE DATA FILE 38,721,970 bp 38,721,980 bp 38,721,990 bp 38,722,000 bp 38,722,010 bp 38,722,020 bp 38,722,030 bp 38,722,040 bp 38,722,050 bp 91 bp chr4 p16.2 p15.33 p15.31 p15.1 p14 p13 p11 q12 q13.1 q13.2 q21.1 q21.23 q22.2 q23 q24 q25 q26 q27 q28.2 q31.1 q31.22 q31.3 q32.2 q33 q34.2 q35.1 IMR90-iPSC coverage IMR90-iPSC [0 - 35] G A Sequence RefSeq genes CpGs CpG islands A G A G T A G C A A A C A C T A A T C T A C G A A T A A T G A A C A T A G G C A T T A T T T T A A G A A C C A A A A G A A A G C A C G T G G G C A T T T G G T T T A C A C A T C A C T CpGs  
71. NAME DATA TYPE DATA FILE 38,721,970 bp 38,721,980 bp 38,721,990 bp 38,722,000 bp 38,722,010 bp 38,722,020 bp 38,722,030 bp 38,722,040 bp 38,722,050 bp 91 bp chr4 p16.2 p15.33 p15.31 p15.1 p14 p13 p11 q12 q13.1 q13.2 q21.1 q21.23 q22.2 q23 q24 q25 q26 q27 q28.2 q31.1 q31.22 q31.3 q32.2 q33 q34.2 q35.1 IMR90-iPSC coverage IMR90-iPSC [0 - 35] G A Sequence RefSeq genes CpGs CpG islands A G A G T A G C A A A C A C T A A T C T A C G A A T A A T G A A C A T A G G C A T T A T T T T A A G A A C C A A A A G A A A G C A C G T G G G C A T T T G G T T T A C A C A T C A C T Second CpG Methylated Unmethylated Total First CpG Methylated 1 2 3 Unmethylated 0 4 4 Total 1 6 7 CpGs  
72. NAME DATA TYPE DATA FILE 38,721,970 bp 38,721,980 bp 38,721,990 bp 38,722,000 bp 38,722,010 bp 38,722,020 bp 38,722,030 bp 38,722,040 bp 38,722,050 bp 91 bp chr4 p16.2 p15.33 p15.31 p15.1 p14 p13 p11 q12 q13.1 q13.2 q21.1 q21.23 q22.2 q23 q24 q25 q26 q27 q28.2 q31.1 q31.22 q31.3 q32.2 q33 q34.2 q35.1 IMR90-iPSC coverage IMR90-iPSC [0 - 35] G A Sequence RefSeq genes CpGs CpG islands A G A G T A G C A A A C A C T A A T C T A C G A A T A A T G A A C A T A G G C A T T A T T T T A A G A A C C A A A A G A A A G C A C G T G G G C A T T T G G T T T A C A C A T C A C T Second CpG Methylated Unmethylated Total First CpG Methylated 1 2 3 Unmethylated 0 4 4 Total 1 6 7 log-odds ratio = log 2 1.5 × 4.5 2.5 × 0.5 ( )=2.4 CpGs  
73. Do  this  50  million  4mes  per  sample   chr strand pos1 pos2 MM MU UM UU! chr1 + 469 471 0 5 1 0! chr1 + 471 484 1 0 4 1! chr1 + 484 489 3 2 1 0! chr1 + 489 493 4 2 1 1! chr1 + 493 497 4 1 3 0! chr1 + 497 525 6 0 1 0! chr1 + 525 542 4 0 0 0! chr1 + 525 563 1 0 0 0! ... ! ... ! ! ... ! !... ! ... ... ... ...!
74. Do  this  50  million  4mes  per  sample  
75. methtuple! www.github.com/PeteHaitch/methtuple  
76. 0 10000 20000 30000 40000 −4 −2 0 2 4 6 8 log2 −odds ratio count ADS: Distance between CpGs = 10 bp
77. −4 −2 0 2 4 6 8 0 10000 20000 30000 40000 count log2 −odds ratio ADS: Distance = 10 bp
78. −4 −2 0 2 4 6 8 0 10000 20000 30000 40000 count log2 −odds ratio ADS: Distance = 10 bp 10% 25% 50% 75% 90% −4 −2 0 2 4 6 8 0 50 100 150 200 250 Distance between CpGs (bp) log2 −odds ratio ADS: Distribution of log2 −odds ratio
79. −4 −2 0 2 4 6 8 0 10000 20000 30000 40000 count log2 −odds ratio ADS: Distance = 10 bp 10% 25% 50% 75% 90% −4 −2 0 2 4 6 8 0 50 100 150 200 250 Distance between CpGs (bp) log2 −odds ratio ADS: Distribution of log2 −odds ratio
80. −4 −2 0 2 4 6 8 0 10000 20000 30000 40000 count log2 −odds ratio ADS: Distance = 10 bp 10% 25% 50% 75% 90% −4 −2 0 2 4 6 8 0 50 100 150 200 250 Distance between CpGs (bp) log2 −odds ratio ADS: Distribution of log2 −odds ratio
81. 10% 25% 50% 75% 90% −4 −2 0 2 4 6 8 0 50 100 150 200 250 Distance between CpGs (bp) log2 −odds ratio ADS: Distribution of log2 −odds ratio
82. 10% 25% 50% 75% 90% −4 −2 0 2 4 6 8 0 50 100 150 200 Distance between CpGs (bp) log2 −odds ratio quantile 10 25 50 75 90 ADS: Distribution of log2 −odds ratio
83. ADS ADS−adipose ADS−iPSC −2 0 2 4 6 0 50 100 150 200 0 50 100 150 200 0 50 100 150 200 Distance between CpGs (bp) log2 −odds ratio quantile 10 25 50 75 90
84. CpG island Non CpG island 0.0 2.5 5.0 0.0 2.5 5.0 0.0 2.5 5.0 ADS ADS−adipose ADS−iPSC 0 50 100 150 200 0 50 100 150 200 Distance between CpGs (bp) log2 −odds ratio CGI CpG island Non CpG island quantile 10 25 50 75 90
85. Co-­‐methyla4on  =  correla4on   “The  rela<onship  between  the  degree  of   methyla<on  over  distance”   1.  Within-­‐fragment  co-­‐methyla4on   2.  Correla/on  of  β-­‐values   Eckhardt,  Florian,  et  al.  "DNA  methyla4on  profiling  of  human  chromosomes  6,  20  and  22."  Nature  gene4cs  38.12  (2006):  1378-­‐1385.  
89. −1.0 −0.5 0.0 0.5 1.0 0 500 1000 1500 Distance between CpGs (bp) Pearson correlation ADS: Correlation of β−values
90. ADS ADS−adipose ADS−iPSC −1.0 −0.5 0.0 0.5 1.0 0 500 1000 1500 0 500 1000 1500 0 500 1000 1500 Distance between CpGs (bp) Pearson correlation
91. ADS ADS−adipose ADS−iPSC −1.0 −0.5 0.0 0.5 1.0 0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500 Distance between CpGs (bp) Pearson correlation
92. ADS ADS−adipose ADS−iPSC −1.0 −0.5 0.0 0.5 1.0 0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500 Distance between CpGs (bp) Pearson correlation "Pioneer  Factor  rearrange  the  nucleosome"  by  Wenqiang  Shi  -­‐  Own  work.  Licensed  under  Crea4ve  Commons  AEribu4on-­‐Share  Alike   3.0  via  Wikimedia  Commons  -­‐  hEp://commons.wikimedia.org/wiki/ File:Pioneer_Factor_rearrange_the_nucleosome.jpg#mediaviewer/File:Pioneer_Factor_rearrange_the_nucleosome.jpg  
93. ADS ADS−adipose ADS−iPSC −1.0 −0.5 0.0 0.5 1.0 0 100 200 300 0 100 200 300 0 100 200 300 Distance between CpGs (bp) Pearson correlation CpG island Non CpG island
94. CGI 1RQï&*, 0 1 2 3 4 0 1 0 1 ` values density data Real (ADS) methsim Distribution of ` values 0 1 0 1 CGI 1RQï&*, 0 250 500 750 1000 Distance between CpGs (bp) Pearson correlation data Real (ADS) methsim Correlations of pairs of ` values 0 4 0 4 CGI 1RQï&*, 0 50 100 150 200 Distance between CpGs (bp) median log odds ratio data Real (ADS) methsim Within haplotype co-methylation at neighbouring CpGs methsim! www.github.com/PeteHaitch/methsim  
95. Methylome  of  the  agou/  viable  yellow   mouse  (Avy)   Some  of  these  mice  are  not  like  the  others  (we  hope…)  
96. Morgan,  Hugh  D.,  et  al.  "Epigene4c  inheritance  at  the  agou4  locus  in  the  mouse."  Nature  gene4cs  23.3  (1999):  314-­‐318.   Methylated  Avy     Unmethylated  Avy    
97. Experimental  design   C57BL/6   Isogenic*  
98. Experimental  design   C57BL/6   Isogenic*   "Liver  (transparent)"  by  Mikael  Häggström  -­‐  File:Human  Hepar.jpg.  Licensed  under  Public  domain  via  Wikimedia  Commons  –     hEp://commons.wikimedia.org/wiki/File:Liver_(transparent).png#mediaviewer/File:Liver_(transparent).png  
99. Experimental  design   C57BL/6   Isogenic*  
100. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a   C57BL/6   Isogenic*  
101. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a   LiEermates  
102. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a   Avy  +  3  Mb  of  C3H  
103. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a  
104. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a   +   30×  whole-­‐genome  bisulfite-­‐sequencing   =     epialleles  
105. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a   +   30×  whole-­‐genome  bisulfite-­‐sequencing   =     $$$$$  
106. Experimental  design   a/a   a/a   a/a   Avy/a   Avy/a   +   30×  whole-­‐genome  bisulfite-­‐sequencing   =     epialleles  
107. Epialleles   a/a   a/a   a/a   Avy/a   Avy/a  
108. Epialleles   a/a   a/a   a/a   Avy/a   Avy/a  
109. Epialleles   a/a   a/a   a/a   Avy/a   Avy/a  
110. Method   1.  Some  of  these  mice  are  not  like  the  others?   2.  Is  my  neighbour  also  different?   3.  Is  my  neighbour  different  in  the  same  way  as   me?  
111. Some  of  these  mice  are  not  like  the  others?   Mouse1   Mouse2   Mouse3   Mouse4   Mouse5   Methylated   17   31   15   23   9   Unmethylated   1   3   0   1   1   P-­‐value  =  0.76  
112. Some  of  these  mice  are  not  like  the  others?   Mouse1   Mouse2   Mouse3   Mouse4   Mouse5   Methylated   38   79   59   69   44   Unmethylated   1   2   1   2   46   P-­‐value  =  2  ×  10-­‐25      
113. Some  of  these  mice  are  not  like  the  others?   P-­‐value  <  threshold   à  (candidate)  differen4ally  methylated  CpG  (DMC)   Mouse1   Mouse2   Mouse3   Mouse4   Mouse5   Methylated   38   79   59   69   44   Unmethylated   1   2   1   2   46   P-­‐value  =  2  ×  10-­‐25      
114. Is  my  neighbour  also  different?  
115. "Run-­‐DMC  Logo”  Licensed  under  Public  domain  via  Wikimedia  Commons  -­‐  hEp://commons.wikimedia.org/ wiki/File:Run-­‐DMC_Logo.svg#mediaviewer/File:Run-­‐DMC_Logo.svg  
116. Is  my  neighbour  also  different?   1.  Find  runs  of  CpGs     –  P-­‐value  <  threshold     –  Within  distance  of  next  CpG   –  Some  allowance  for  missing  or  “insignificant”  CpGs   2.  Filter  candidate  runs   –  Run  contains  enough  CpGs  
117. Is  my  neighbour  different  in  the  same   way  as  me?  
118. 0.00 0.25 0.50 0.75 1.00 1000 1100 1200 1300 start β sample mouse1 mouse2 mouse3 mouse4 mouse5 An inconsistent candidate region Data:  Emma  Whitelaw  
119. 0.00 0.25 0.50 0.75 1.00 1000 1100 1200 1300 start β sample mouse1 mouse2 mouse3 mouse4 mouse5 An inconsistent candidate region Data:  Emma  Whitelaw  
120. Is  my  neighbour  different  in  the  same   way  as  me?   •  Flag  regions  with  3-­‐way  interac4on  between   sample  ×  methyla4on  level  ×  posi4on   – Not  quite  what  we  want   – So  plot,  plot,  plot    
121. CpG island gene 0.0 0.5 1.0 1000 2000 3000 4000 Position (bp) β−value Cov 0 25 50 75 sample mouse1 mouse2 mouse3 mouse4 mouse5 Data:  Emma  Whitelaw  
122. Summary   How  I  work,  what  I  found,  and  what  I’m  proud  of  
123. “Doesn’t  the  gardener  lavish  more  care  on   the  thorns  than  on  the  flowers”   "Agnon"  of  Unknown  -­‐  The  David  B.  Keidan  Collec4on  of  Digital  Images  from  the  Central  Zionist  Archives  (via  Harvard  University  Library).   Licensed  under  the  Public  Domain  via  Wikimedia  Commons  -­‐  hEp://commons.wikimedia.org/wiki/File:Agnon.jpg#mediaviewer/File:Agnon.jpg   -­‐  Hartman  in  Metamorphisis  by  S.Y.  Agnon   Via  @erichlya  
124. “You  can  observe  a  lot  by  watching”   "Yogi  Berra  1956"  by  unknown  -­‐  Baseball  Digest,  front  cover,  September  1956  issue.  [1].  Licensed  under  Public  domain  via  Wikimedia  Commons   -­‐  hEp://commons.wikimedia.org/wiki/File:Yogi_Berra_1956.png#mediaviewer/File:Yogi_Berra_1956.png   -­‐  Yogi  Berra  
125. -­‐  Robert  Van  Winkle   a.k.a.  Vanilla  Ice   hEp://4.bp.blogspot.com/-­‐N-­‐caq8YoBaQ/UpPZs5EChnI/AAAAAAAABIw/Pb8g7jjJJGA/s1600/Stop+Collaborate.jpg  
126. -­‐  Robert  Van  Winkle   a.k.a.  Vanilla  Ice   hEp://4.bp.blogspot.com/-­‐N-­‐caq8YoBaQ/UpPZs5EChnI/AAAAAAAABIw/Pb8g7jjJJGA/s1600/Stop+Collaborate.jpg  
127. What  I  found   •  Es4mated  strong  spa4al  dependence  of  DNA   methyla4on   •  Cell-­‐type  differences  in  dependence  structure   •  Evidence  of  higher  order  chroma4n  structure   in  spa4al  dependence  data  
128. What  I  found   •  Es4mated  strong  spa4al  dependence  of  DNA   methyla4on   •  Cell-­‐type  differences  in  dependence  structure   •  Evidence  of  higher  order  chroma4n  structure   in  spa4al  dependence  data  
129. What  I  found   •  Es4mated  strong  spa4al  dependence  of  DNA   methyla4on   •  Cell-­‐type  differences  in  dependence  structure   •  Evidence  of  higher  order  chroma4n  structure   in  spa4al  dependence  data  
130. What  I’m  proud  of   www.github.com/PeteHaitch  
131. Acknowledgements   Terry  Speed   Peter  Hall  
132. Acknowledgements   Data   –  Ryan  Lister  et  al.  (UWA,  Salk   Ins4tute)     –  Sue  Clark,  Aaron  Statham  (Garvan   Ins4tute)   –  Emma  Whitelaw,  Harry  Oey  (La   Trobe)   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Everyone  who  makes  their  data   publicly  available   Methodology  &  technology   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Felix  Krueger  (Babraham  Ins4tute)   –  Toby  Sargeant  (WEHI)   –  Keith  SaEerley  (WEHI)   –  Bioconductor  developers   –  WEHI  Bioinforma4cs   –  Everyone  who  makes  their   soSware  open  source   Sanity:  Family  and  friends   Funding:  APA  and  VLSCI   Sanity:  Family  and  friends  
133. Acknowledgements   Data   –  Ryan  Lister  et  al.  (UWA,  Salk   Ins4tute)     –  Sue  Clark,  Aaron  Statham  (Garvan   Ins4tute)   –  Emma  Whitelaw,  Harry  Oey  (La   Trobe)   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Everyone  who  makes  their  data   publicly  available   Methodology  &  technology   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Felix  Krueger  (Babraham  Ins4tute)   –  Toby  Sargeant  (WEHI)   –  Keith  SaEerley  (WEHI)   –  Bioconductor  developers   –  WEHI  Bioinforma4cs   –  Everyone  who  makes  their   soSware  open  source   Sanity:  Family  and  friends   Funding:  APA  and  VLSCI   Sanity:  Family  and  friends  
134. Acknowledgements   Data   –  Ryan  Lister  et  al.  (UWA,  Salk   Ins4tute)     –  Sue  Clark,  Aaron  Statham  (Garvan   Ins4tute)   –  Emma  Whitelaw,  Harry  Oey  (La   Trobe)   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Everyone  who  makes  their  data   publicly  available   Methodology  &  technology   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Felix  Krueger  (Babraham  Ins4tute)   –  Toby  Sargeant  (WEHI)   –  Keith  SaEerley  (WEHI)   –  Bioconductor  developers   –  WEHI  Bioinforma4cs   –  Everyone  who  makes  their   soSware  open  source   Sanity:  Family  and  friends   Funding:  APA  and  VLSCI   Sanity:  Family  and  friends  
135. Acknowledgements   Data   –  Ryan  Lister  et  al.  (UWA,  Salk   Ins4tute)     –  Sue  Clark,  Aaron  Statham  (Garvan   Ins4tute)   –  Emma  Whitelaw,  Harry  Oey  (La   Trobe)   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Everyone  who  makes  their  data   publicly  available   Methodology  &  technology   –  Kasper  Hansen,  Rafael  Irizarry   (Johns  Hopkins,  Harvard)   –  Felix  Krueger  (Babraham  Ins4tute)   –  Toby  Sargeant  (WEHI)   –  Keith  SaEerley  (WEHI)   –  Bioconductor  developers   –  WEHI  Bioinforma4cs   –  Everyone  who  makes  their   soSware  open  source   Sanity:  Family  and  friends   Funding:  APA  and  VLSCI   Sanity:  Family  and  friends  
136. hEp://s18.photobucket.com/user/endriquelimones/media/tensographics_math.jpg.html   Biology   Sta4s4cs   Bioinforma4cs   www.peterhickey.org/presenta4ons