Genome Sequencing and Annotation

1
Genome Sequencing and Annotation
2

• The first objective of most genome projects is to
  determine the DNA sequence either of the genome
  or of a large number of transcripts.

3
Automated DNA Sequencing

• The Principle of Sanger Sequencing

4
ddNTP
5
(No Transcript)
6
Isotope label 33P or 35S
7

• With a separation of 0.5 mm between bands
  corresponding to each molecule that differs in
  length by one nucleotide.
• In general, only up to 500 bases could be read
  from a single set of four lanes on a 30-cm gel.
• Its difficult to sequencing in GC-rich region.

8
High-throughput Sequencing
9

• Four-color fluorescent dyes have replaced the
  radioactive label.
• Rather than stopping the electrophoresis at a
  particular time, the products are scanned for
  laser-induced fluorescence just before they run
  off the end of the electrophoresis medium.
• Reads greater than 1200 bp in a run are possible
  with current technology, though the 500-900 bp
  range is more common.
• Capillary electrophoresis

10
Anion carrier
11
ABI 3700 Sequencer
Twenty 96-well plats are handle per day 2Mb of
sequence in length be determined.
12
Reading Sequence Traces

• The reading of raw sequence traces, or
  base-calling, is now routinely performed using
  automated software that read bases, aligns
  similar sequences, and provides an intuitive
  platform for editing.
• phred free software
• http//www.phrap.org/

13
(No Transcript)
14
2.2 The phred base-calling algorithm
15
2.3 Automated sequence chromatograms
lt 50 bp
5 ? 3
gt 800 bp
16
Contig Assembly
Optimizing sequence alignment with highest score
17
Box 2.1 Pairwise Sequence Alignment
18
2.5 An aligned-reads window in consed
Phrap assembler and consed graphic editor
http//www.phrap.org/
19
Several sequencing technology have been proposed
to reduce the cost.
20
Emerging Sequencing Methods

• Sequencing by hybridization (SBH)
• Mass spectrophotometric techniques
• Nanopore sequencing strategies
• Single-molecule Sanger sequencing
• Polony sequencing

21
(No Transcript)
22
Mass spectrophotometric techniques
23
Mass spectrophotometric techniques DNA
sequencing use the same prinple
24
Nanopore sequencing strategies
25
http//arep.med.harvard.edu/Polonator/chem/
A
C
G
U
26
2.6 Single-molecule polony sequencing
http//arep.med.harvard.edu/Polonator/
27
Genome Sequencing

• Whole chromosome sequences are reassembled from
  the sequences of hundreds of thousands of
  fragments, each typically between 500 and 1000 bp
  in length.
• Two general strategies
• 1. Hierarchical sequencing
• 2. Shotgun sequencing

28
2.7 (Part 1) Hierarchical versus shotgun
sequencing
29
2.7 (Part 2) Hierarchical versus shotgun
sequencing
30
Hierarchical sequencing

• Also known as top-down, map-based, or
  clone-by-clone strategy, developed in the late
  1980s.
• The first step is to clone the genome as
  manageable units of some 50-200 kb in length,
  which relative locations are known.
• DNA libraries are constructed by partial
  digestion or shearing of genomic DNA and are
  ligated into cloning site using standard
  recombinant DNA procedurce.

31
The copy number close to one per cell
32
2.8 Cloning vectors used in genome sequencing
33
2.9 Hierarchical assembly of a sequence-contig
scaffold (supercontig) --- tiling path
5- to 10-fold redundancy

• Assembly methods
• Hybridization (A)
• Fingerprinting
• (B,C,D)
• 3. End-sequencing

34
2.10 Aligning BAC clones by hybridization and
fingerprinting
35
End-sequencing
36
Shotgun sequencing

• Sequence fragments with 5- to 10-fold redundancy
  are generated from a plasmid library that
  constructed from a single whole genome.
• The contig scaffold are produced by Celera
  Genomics using computer algorithm such as
  Screener, Overlapper, and Unitigger.

37
2.11 U-unitigs and repeat resolution
Overlaps of at least 40 bp with no more 6
differences were accepted.
38
2.12 Proportion of fly and human genomes in
large scaffolds
39
Sequence Verification

• Completeness
• Accuracy
• Validity of assembly

40
2.13 Alignment of two draft human genome
assemblies
41
Genome Annotation
42
EST Sequencing
43
EST
44
2.14 (Part 1) Relationship between gene
structure, cDNA, and EST sequences
Alternative splicing
45

• Alternative splicing of mRNA
• Splice pair

GU donor site
AG acceptor site
Exon Enhancer Motif
46

• Four Modes of Alternative Splicing

Splice / Don't Splice   
Competing 5' or 3' Splice Sites 
47
Exon Skipping  
Mutually Exclusive Exons 
48
Ab Initio Gene Discovery
To predict gene according to sequence pattern
Flanking region
Flanking region
5'
3'

GT
AG
GT
AG
GC box
Initiation codon
Stop codon
Poly(A)-addition site
CAAT box
TSS
AATAA
GC box
TATA box
49

• Statistics-based (ab initio) Methods
• Ab initio approach predict genes directly
  using the computational properties of exons,
  introns, and other features in the genomic
  sequences without the reference of the
  experimental data.
• (1) Hidden Markov Model
• GenScan, Genie, Genemark, Veil,
  HMMgene
• (2) Neural Networks
• Grail II, GrailEXP_Perceval
• (3) Decision Tree
• MZEF, MZEF-SPC
• (4) Integration of Various Statistical
  Approaches
• FGENESH

50
Example Two dices1
S1
S2

• One dice is fair, producing 1,2,3,4,5,6 with
  equal probability. The other one is loaded and
  produce number 6 with higher probability than
  the rest numbers.
• Each time only one dice is rolled and produces
  numbers in the set 1,2,3,4,5,6 which can be
  observed. But we do not know which of the two
  dices are rolled.
• The transitions between the fair dice and loaded
  dice follow the Markov chain properties.

0.9
0.9
0.1
fair
loaded
0.1
1 1/6 2 1/6 3 1/6 4 1/6 5 1/6 6 1/6
1 1/10 2 1/10 3 1/10 4 1/10 5 1/10 6 1/2
6 4 5 3 1 3 2 6 5 1 4 5 6 3 6 6 6 4 6 3 1 6 5 6 6
6 2 4 5 3
F F F F F F F F F F F F L L L L L L L L L L L L L
L F F F F
51
Elements of a hidden markov model

• An HMM is characterized by the follow elements
• N, the number of hidden states. S1, S2, SN
• ( N 2 Fair dice, loaded dice )
• M, the number of distinct observations. V1, V2,
  VM
• (M 6 1, 2, 3, 4, 5, 6)
• Transition probabilities between the hidden
  states
• aij P(Sj Si is the previous state) for each
  i, j from 1 to N.
• ( P(FF) 0.9, P(FL)0.1, P(LL) 0.9, P(LF)
  0.1 )
• Emission probability of each observation for each
  hidden state bi(k) P(VkSi)
• (bFair(k) 1/6 for k 1,2,3,4,5,6
  bLoaded(k)1/10 for k 1,2,3,4,5 and bLoaded(6)
  1/2)
• Initial probabilities probability of each state
  being the first state.

52
Modeling Exon-Intron Segments

• We can treat the exon and intron segments on DNA
  sequences as two distinct hidden states, each of
  which emits A, T, C, G with its own
  probabilities.

S1
S2
0.9
0.9
0.1
exon
intron
0.1
A 1/6 T 1/6 C 1/3 G 1/3
A 1/4 T 1/4 C 1/4 G 1/4
53
Modeling Exon-Intron Segments

• Number of hidden states 2 (exon state and intron
  state)
• Number of observations 4 (A, T, C, G)
• Transition probabilities
• P(S1?S2) 0.1
• P(S1?S1) 0.9
• P(S2?S1) 0.1
• P(S2?S2) 0.9
• Emission probabilities
• S1 P(A)P(T) 1/6
• P(C)P(G) 1/3
• S2 P(A)P(T)P(C)P(G)1/4

S1
S2
0.9
0.9
0.1
exon
intron
0.1
A 1/6 T 1/6 C 1/3 G 1/3
A 1/4 T 1/4 C 1/4 G 1/4
54
Neural Networks
55

• Homology-based Methods
• Homological approach identifies genes
  with the aid of experimental data. This approach
  exploits the alignment gene sequence between
  genomic data and the known cDNA (or protein)
  database.
• (1) Local Alignment Methods (BLAST
  -based)
• AAT, GAIA , INFO, TAP
• (2) Pattern-based Alignment Method
• Flash, ICE

56

• Combination Tools
• These tools combine both sequence similarity
  and ab initio gene finding approaches. They
  predict genes by producing a splicing alignment
  between a genomic sequence and a candidate amino
  acid sequence.
• Procrustes, GeneWise, GenomeScan
• FGENESH and FGENESH
• GrailEXP_Gawain and _GALAHAD

57
Evaluate accuracy

• Definition

TPP
TPE
FP
FN
TN
TP(True Positive) correctly predicted as
coding TN(True Negative) correctly predicted as
noncoding FP(False Positive) noncodingà
coding FN(False Negative) coding à noncoding
58

• Sensitivity
• SnTPE/(TPFN)
• Specificity
• SpTPE/(TPFP)
• Missing exon
• ME (number of missing exons)/(number
  of actual exons)
• Wrong exon
• WE(number of wrong exons)/(number of
  predicted exons)

59
Accuracy of exon prediction

• Statistics-based (ab initio) Methods
• Sensitity/Specificity 30 70
• Homology-based Methods
• Sensitity/Specificity 99
• when related protein sequences available
• Combination Tools
• ????????

60
Regulatory Sequences
Phylogenetic footprinting two sequence
alignment and conserved pattern discovery, in
principle.
61
2.15 Phylogenetic shadowing
Uncover the conserved segments with multiple
sequence alignment
Candidate of Regulatory sequence
Larger diversity, Non-regulatory sequence
62
Non-Protein Coding Genes

• It is difficult to identify non-protein coding
  genes
• Most of the transcripts are not polyadenylated,
  and so lack of cDNA libraries.
• Conserved in the level of secondary structure,
  poor sequence constraint.
• Relatively little is known about the function and
  distribution.

63
Structural Features of Genome Sequences

• Repetitive sequences (five classes)
• GC content
• Simple sequence repeats
• Segmental duplication
• Structure of centromeres and telomeres

64
The ENCODE Project

• Only 5 of mammalian genome sequence is highly
  conserved.
• At most, only 2 of a typical mammalian genome
  encode transcripts.
• Most of the non-genic sequences are unknown.

65
(No Transcript)
66
Clusters of Orthologous Genes (COGs)
paralog
ortholog
67
http//www.ncbi.nlm.nih.gov/COG/
68
Gene Ontology
69
(No Transcript)
70
http//www.geneontology.org/