2. Lecture 1 Introduction to high throughput sequencing
Michael Brudno
CSC 2431
January 13, 2010
Adapted from presentations by Francis Ouelette, OICR, Michael Stromberg, BC and Asim Siddiqui, ABI

3. DNA sequencing How we obtain the sequence of nucleotides of a species
…ACGTGACTGAGGACCGTG
CGACTGAGACTGACTGGGT
CTAGCTAGACTACGTTTTA
TATATATATACGTCGTCGT
ACTGATGACTAGATTACAG
ACTGATTTAGATACCTGAC
TGATTTTAAAAAAATATT…

4. DNA Sequencing Goal: Find the complete sequence of A, C, G, T’s in DNA
Challenge:
There is no machine that takes long DNA as an input, and gives the complete sequence as output
Can only sequence ~500 letters at a time

5. Generations of Sequences
Sanger-style: Classic
454 “First Next-gen”
Illumina + ABI SOLiD “Next-gen”
Helicos “2.5 Gen”
PacBio “Next-next-gen”, 3rd gen

6. After Next-generation:
Why are we sequencing?
Before Next-generation:
DNA, RNA, (proteins), (populations), sampling, averages, consensus
Problems: sampling, averages, consensus.
After Next-generation:
Genome sequence and structure
Less cloning/PCR
Single molecules (for some)

7. Sanger (old-gen) Sequencing
Now-Gen Sequencing
Whole Genome
Human (early drafts), model organisms, bacteria, viruses and mitochondria (chloroplast), low coverage
New human (!), individual genome, 1,000 normal, 25,000 cancer matched control pairs, rare-samples
RNA
cDNA clones, ESTs, Full Length Insert cDNAs, other RNAs
RNA-Seq: Digitization of transcriptome, alternative splicing events, miRNA
Communities
Environmental sampling, 16S RNA populations, ocean sampling,
Human microbiome, deep environmental sequencing, Bar-Seq
Other
Epigenome, rearrangements, ChIP-Seq

8. Differences between the various platforms:
Nanotechnology used.
Resolution of the image analysis.
Chemistry and enzymology.
Signal to noise detection in the software
Software/images/file size/pipeline
Cost $$$

9. Next Generation DNA Sequencing Technologies
Adapted from Richard Wilson, School of Medicine, Washington University, “Sequencing the Cancer Genome”
Next Generation DNA Sequencing Technologies
Human Genome
6GB == 6000 MB
Req’d Coverage 6 12 30
3730
454
Illumina
bp/read
600
400
2X75
reads/run 96
500,000
100,
bp/run
57,600
0.5 GB
15 GB
# runs req’d
625,000
144
runs/day 2 1
0.1
Machine days/human genome
312,500 (856 years)
120
Cost/run
$48
$6,800
$9,300
Total cost
$15,000,000
$979,200
$111,600

10. Solexa-based Whole Genome Sequencing
Adapted from Richard Wilson, School of Medicine, Washington University, “Sequencing the Cancer Genome”

11. Illumina (Solexa)

12. Illumina (Solexa)

13. Illumina (Solexa)

14. From Debbie Nickerson, Department of Genome Sciences, University of Washington,

15. What is a base quality? Base Quality Perror(obs. base) 3 50.12% 5
31.62% 10
10.00% 15
3.16% 20
1.00% 25
0.32% 30
0.10% 35
0.03% 40
0.01%

16. Next-gen sequencers From John McPherson, OICR bases per machine run
100 Gb
AB/SOLiDv3, Illumina/GAII
short-read sequencers
(10+Gb in bp reads,
>100M reads, 4-8 days)
10 Gb
1 Gb
454 GS FLX pyrosequencer
( Mb in bp reads,
0.5-1M reads, 5-10 hours)
bases per machine run
100 Mb
10 Mb
ABI capillary sequencer
Very different types of data. Different run times. Different costs
( Mb in bp reads,
96 reads, 1-3 hours)
1 Mb
10 bp
100 bp
1,000 bp
read length

17. DNA sequencing – vectors
Shake
DNA fragments
Known
location
(restriction
site)
Vector
Circular genome
(bacterium, plasmid) + =

18. Method to sequence longer regions
genomic segment
cut many times at random (Shotgun)
Get two reads from each segment
~500 bp
~500 bp

19. Reconstructing the Sequence (Fragment Assembly)
reads
Cover region with ~7-fold redundancy (7X)
Overlap reads and extend to reconstruct the original genomic region

20. Definition of Coverage
Length of genomic segment: L
Number of reads: n
Length of each read: l
Definition: Coverage C = n l / L
How much coverage is enough?
Lander-Waterman model:
Assuming uniform distribution of reads, C=10 results in 1 gapped region /1,000,000 nucleotides

21. Challenges with Fragment Assembly
Sequencing errors
~1-2% of bases are wrong
Repeats
Computation: ~ O( N2 ) where N = # reads
false overlap due to repeat

22. History of DNA Sequencing
Adapted from Eric Green, NIH; Adapted from Messing & Llaca, PNAS (1998)
1870
Miescher: Discovers DNA
Avery: Proposes DNA as ‘Genetic Material’
Efficiency
(bp/person/year)
1940
Watson & Crick: Double Helix Structure of DNA
1953 1
Holley: Sequences Yeast tRNAAla 15
1965
Wu: Sequences  Cohesive End DNA
150
1970
Sanger: Dideoxy Chain Termination
Gilbert: Chemical Degradation
1,500
1977
15,000
Messing: M13 Cloning
1980
25,000
Hood et al.: Partial Automation
50,000
1986
Cycle Sequencing
Improved Sequencing Enzymes
Improved Fluorescent Detection Schemes
200,000
1990
50,000,000
2002
Next Generation Sequencing
Improved enzymes and chemistry
New image processing
100,000,000,000
2009

23. Which representative of the species?
Which human?
Answer one:
Answer two: it doesn’t matter
Polymorphism rate: number of letter changes between two different members of a species
Humans: ~1/1,000 – 1/10,000
Other organisms have much higher polymorphism rates
Just beneath the surface of the ocean floats a tiny egg. Within it's hull of follicle cells the embryo of Ciona, a sea squirt or Ascidian, is beginning to develop. The shape of the egg does not give any clue of the creature that it is going to be. Now about a third of a millimeter it is waiting to become a larva, and this larva is a rather surprising creature.
The larva is developing within a few weeks. A head and a tail are evolving. When you look carefully you can see a rod that stiffens the tail. We know such a device as a notochord. A trained zoologist would identify the animal as belonging to the Phylum Chordata, and that is the same group we humans belong to. (See footnote).
Also the internal organs are beginning to show. But the image is a bit deceiving. What seems to be an eye is in fact a device for equilibrium called the 'Otolith'. Above it we find the light sense organ, the 'Ocellus'
So there is something suspicious about these so-called sea squirts. Freed from it's shell a familiar form has appeared. Clearly the resemblance with a tadpole larva is seen. The little creature swims for a few hours to find a good spot somewhere on a solid surface. Then a surprising thing happens.
This larva is not going to evolve into a fish, amphibian or anything like that. With the front of it's head it attaches itself to a surface. Within minutes resorption of the larva tail commences and the sea squirt will stay on that same spot for all it's life

24. Why humans are so similar
A small population that interbred reduced the genetic variation
Out of Africa ~ 40,000 years ago
Out of Africa

25. Migration of human variation

26. Migration of human variation

27. Migration of human variation

28. Genetic Variations: Why?
Phenotypic differences
Inherited diseases
Ancestral history

29. Genetic Variations: SNPs & INDELs

30. Structural Variations
Paul Medvedev
review in prep
July 2009

31. SNP Discovery: Goal
sequencing errors
SNP

32. SNP Discovery: Base Qualities
High quality
Low quality

33. SNPs & Bayesian Statistics
# of individuals
base quality
allele call in read

34. SNP Discovery haploid diploid AACGTTAGCATA AACGTTAGCATA strain 1
AACGTTCGCATA
strain 1
individual 1
AACGTTCGCATA
strain 2
AACGTTCGCATA
AACGTTAGCATA
individual 2
strain 3
AACGTTAGCATA
individual 3

35. Genotyping & Consensus Generation
haploid
diploid
AACGTTAGCATA
AACGTTAGCATA
AACGTTCGCATA
strain 1
[A]
individual 1
[A/C]
strain 2
[C]
AACGTTCGCATA
AACGTTCGCATA
individual 2
[C/C]
AACGTTAGCATA
strain 3
[A]
individual 3
[A/A]
AACGTTAGCATA

36. Visualization: Consed

37. 1000 Genomes Project

38. 1000G: Goals Discover genetic variations Variant alleles
1 % minor allele frequencies across genome
0.1 – 0.5 % MAF across gene regions
Variant alleles
Estimate frequencies
Identify haplotype background
Characterize linkage disequilibrium

39. 1000G: Pilot Projects Pilot 1 Pilot 2 Pilot 3 Low coverage 180 samples
70 4X
110 2X
2.7 Tbp total
202 Gbp 454
1.8 Tbp Illumina
640 Gbp AB SOLiD
Pilot 2
Deep trios (CEU & YRI)
6 samples
1.1 Tbp total
87 Gbp 454
773 Gbp Illumina
270 Gbp AB SOLiD
Pilot 3
Exon capture
607 samples
2.2 Mbp of targets
8800 targets
10 – 20x coverage

40. Questions about the genome
Obtaining a genome sequence is a one step towards understanding biological processes
Questions that follow from the genome are:
What is transcribed?
Where do proteins bind?
What is methylated?
In other words, how does it work?

41. Central dogma ZOOM IN tRNA transcription DNA rRNA snRNA translation
POLYPEPTIDE
mRNA

42. Transcription The DNA is contained in the nucleus of the cell.
A stretch of it unwinds there, and its message (or sequence) is copied onto a molecule of mRNA.
The mRNA then exits from the cell nucleus.

43. DNA
RNA T C A G G A T C G A U C
A = T
G = C
T  U

44. More complexity The RNA message is sometimes “edited”.
Exons are nucleotide segments whose codons will be expressed.
Introns are intervening segments (genetic gibberish) that are snipped out.
Exons are spliced together to form mRNA.

45. Splicing frgjjthissentencehjfmkcontainsjunkelm
thissentencecontainsjunk

46. Key player: RNA polymerase
It is the enzyme that brings about transcription by going down the line, pairing mRNA nucleotides with their DNA counterparts.

47. Promoters
Promoters are sequences in the DNA just upstream of transcripts that define the sites of initiation.
The role of the promoter is to attract RNA polymerase to the correct start site so transcription can be initiated. 5’ 3’
Promoter

48. Promoters
Promoters are sequences in the DNA just upstream of transcripts that define the sites of initiation.
The role of the promoter is to attract RNA polymerase to the correct start site so transcription can be initiated. 5’ 3’
Promoter

49. Transcription – key steps
DNA
Initiation
Elongation
Termination
DNA +
RNA

50. Genes can be switched on/off
In an adult multicellular organism, there is a wide variety of cell types seen in the adult. eg, muscle, nerve and blood cells.
The different cell types contain the same DNA though.
This differentiation arises because different cell types express different genes.
Promoters are one type of gene regulators

51. Transcription (recap)
The DNA is contained in the nucleus of the cell.
A stretch of it unwinds there, and its message (or sequence) is copied onto a molecule of mRNA.
The mRNA then exits from the cell nucleus.
Its destination is a molecular workbench in the cytoplasm, a structure called a ribosome.

52. The Transcriptome
The transcriptome is the entire set of RNA transcripts in the cell, tissue or organ.
The transcriptome is cell type specific and time dependant i.e. It is a function of cell state
The transcriptome can help us understand how cells differentiate and respond to changes in their environment.

53. Transcriptome complexity
Transcripts may be:
Modified
Spliced
Edited
Degraded
Transcriptome is substantially more complex than the genome and is time variant.

54. ESTs ESTs were the first genome wide scan for transcriptional elements
Different library types:
Proportional
Normalized
Subtractive
Can be sequenced from the 5’ or 3’ end

55. “Hello Mr Chips” Microarray chips introduced in 90’s
Parallel way to measure many genes
Probes placed on slides
RNA -> cDNA, labelled with fluorescent dye and hybridized.
Fluorescence measured
Chips have been highly successful
Simplified analysis
Useful when there is no genome sequence
Linear signal across 500 fold variation
Standardization has aided use in medical diagnostics
E.g. Mammaprint

56. Microarray expression profiling by 2-color assay (“cDNA arrays”)
PCR products
6250 yeast ORFs
hybridized cDNAs:
green = control
red = experiment
*Schena et al., 1995

57. Chips: pros and cons Advantages Disadvantages
Do not require a genome sequence
Highly characterised, with many s/w packages available
One Affymetrix chip FDA approved
Disadvantages
Measurements limited to what’s on the array
Hard to distinguish isoforms when used for expression
Can’t detect balanced translocations or inversions when used for resequencing

58. mRNA-seq Basic work flow
Align reads (sometimes to transcriptome first and then the genome)
Tally transcript counts
Align tags to spliced transcripts
Add to transcript counts

59. Cloonan et al
Used SOLiD to generate 10Gb of data from mouse embryonic stem cells and embryonic bodies
Used a library of exon junctions to map across known splice events

60. Distribution of tags

61. Tag locations

62. General issues Coverage across the transcript may not be random
Some reads map to multiple locations
Some reads don’t map at all
Reads mapping outside of known exons may represent
New gene models
New genes

63. Size of the transcriptome
Carter et al (2005)
Using arrays estimated 520,000 to 850,000 transcripts per cell.
Use upper limit and estimate average transcript size of 2kb
Transcriptome ~2GB
Transcriptome cost ~ genome cost

64. The Boundome DNA binding proteins control genome function
Histones impact chromatin structure
Activators and repressors impact gene expression
The location of these proteins helps us understand how the genome works

65. ChIP

66. Chip-Seq Instead of probing against a chip, measure directly
Basic work flow
Align reads to the genome
Identify clusters and peaks
Determine bound sites

67. Robertson et al. 2007
Used Illumina technology to find STAT1 binding sites
Comparisons with two ChIP-PCR data sets suggested that ChIP-seq sensitivity was between 70% and 92% and specificity was at least 95%.

68. Tag statistics

69. Typical Profile

70. Mikkelsen et al., 2007
Performed a comparison with ChIP-chip methods ~98% concordance

71. Comparison with ChIP-seq

72. The Methylome In methylated DNA, cytosines are methylated.
This leads to silencing of genes in the region e.g. X inactivation
It is yet another form of transcriptional control and together with histone modifications a key component of epigenetics

73. Bi-sulphite sequencing
Converts un-methylated cytosines to uracil (which becomes thymine when converted to cDNA)
Experimental procedure is difficult
Sequence alignment is tricky, but the basic concepts hold

74. Taylor et al, 2007 Targeted sequencing reduced alignment difficulties
Used dynamic programming to identify alignments of sequences against an in silico bisulphate converted sequence of the target amplicon regions

75. Metagenomics
Craig Venter’s sequencing of the sea one of the earliest and most well known examples
Used Sanger sequencing
Many recent studies including
Angly et al – studied ocean virome
Cox-Foster et al – studied colony collapse disorder
All use 454 for its longer read length and target amplification of 16S or 18S ribsomal subunits