Alignment of whole genomes using suffix trees

1
Alignment of whole genomes using suffix trees

• Mahshid Shakiba
• Nov 17, 2004

IFT 6299, University of Montreal
2
Outline

• Motivation
• MUMmer
• Algorithms
• Observations
• MUMmer 2
• Algorithms
• NUCmer
• PROmer
• Observations
• MUMmer 3

3
Motivation

• We need to
• Align the entire genomes of two closely related
  organisms (millions bps)
• Compare sequence assemblies at different stages
  of project
• Compare the results of different assembly
  algorithms
• Current algorithms
• Ineffective in aligning very long sequences of
  DNA
• Can not detect large scale changes

4
Application

• MUMmer , a system to align and compare very large
  DNA Protein sequences in linear time
• History
• MUMmer 1.0, 1999
• MUMmer 2.1, 2002
• MUMmer 3.0, 2004
• Created at Tiger institute
• Website http//www.tigr.org/software/mummer/

5
MUMmer 1.0

• Assumption Two DNA sequences are closely
  related
• Inputs Two DNA sequences and length of the
  shortest MUM (Maximal Unique Match)
• Output A base-to-base alignment of input
  sequences, highlights SNPs, large inserts,
  significant repeats ,transpositions and reversals
  
• Techniques Suffix trees , LIS (Longest
  Increasing Subsequences), Smith-Waterman
  alignment algorithm

6
MUMmer

• Algorithm

1- Construct suffix tree for AB 2- Finding,
sorting MUMs 3-Matching MUMs 4-Closing gaps
MUMs
gaps
7
Suffix tree

• Definition a compact representation of all
  possible suffixes of an input S
• Can be built in O(m) time and space where m S
• Search of sub-string X takes O(n) time, n X

8
Suffix tree
Suffixes 1 gaaccgacct 2 aaccgacct 3
accgacct 4 ccgacct 5 cgacct 6
gacct 7 acct 8
cct 9 ct 10
t
9
Maximal Unique Match

• Sequences in genomes A and B that
• occur exactly once in A and in B
• are not contained in any larger such sequence
• Genome A tcgatcGACGATCGCAGCATAAcgact
• Genome B gcattaGACGATCGCAGCATAAtcca

A
B
10
Finding, sorting MUMs

• MUM Internal node with a leaf from each genome
  in its subtree
• With single scan of the suffix tree, find all
  MUMs
• Sort MUMs based on their position in genome A.

11
Matching MUMs
2
1
3
4
5
6
7
A
B
3
1
2
4
5
7
6
Select longest consistent set of MUMs occur in
the same order in A and B
2
1
4
5
7
A
B
1
2
4
5
7
12
Closing Gaps

• Gaps interruption in MUM-alignment
• Types of gaps
• SNP Single Nucleotide Polymorphisms
• Insertion
• Highly polymorphic region
• Repeat
• Closing methods
• Repeat procedure using a shorter minimum length
  for MUMS (long gaps)
• Use the Smith-Waterman alignments ( short gaps)

13
MUMmer Output
14
Observations

• Align 2 highly homologous strains of
  M.tuberculosis, 4.4 million bps.
• Time 5 s suffix tree construction, 45 s sorting
  MUMs, 5 s Smith-Waterman alignments.
• Longest MUM ,24 563 bp 249 MUMs gt 5000 bp gt90
  identical
• Align 2 cousin bacteria, M.genitalium (580 kbp)
  and M.pneumoniae (816 kbp)
• Time 6.5 s suffix tree finding LIS 0.02 s 116
  s alignments.
• Longest MUM, 281 bp, 16 MUMs gt 100 bp, lt50
  identical

15
Observations

• Align 2 syntenic sequences from human chromosome
  12 and mouse chromosome 6 (225 kbp).
• Time 29 s in total, 1.6 s for suffix tree.
• Longest MUM, 117 bp, 10 MUMs gt 50bp

16
MUMmer 2

• Problem with MUMmer 1
• Align only DNA sequences
• Needs lots of memory
• Can not align incomplete genomes
• Solution MUMmer2
• 3 times faster than MUMmer 1
• Requires 1/3 space
• Can align protein strands and incomplete genomes
• Parallel alignment

17
MUMmer 2

• Algorithm
• Use only 20 bytes per bp (MUMmer, 38 bytes)
• Build suffix tree for the shorter sequence
• Find MUMs by streaming the second sequences
  against suffix tree
• cluster the matches

18
Streaming algorithm
Streaming String
...atgtcc...
atgtgtgtc

c
t
gt
9
1
10
i
1
c
c
gt
gt
8
7
i
c
gtc
c
gt
5
3
6
Suffix Tree for String
atgtgtgtc
1
2
3
4
5
6
7
9
10
c
gtc
8
Use suffix links to find the start of next match.
4
2
19
Cluster MUMs

• Align unfinished assembly which needs
  rearrangement
• Cluster MUMs which are
• Close enough
• Find Longest Increasing Subsequence

20
NUCmer

• Multiple-contigs alignment program
• Uses MUMmer 2
• Can
• Compare assemblies at different stages of project
• Compare unfinished genomes to a closely related
  genome (speed up finishing step)
• Compare outputs of 2 different assembly program

21
NUCmer

• Inputs 2 multi-fasta files
• Output alignment of every contig in the first
  file to every sequence in the second file
• Algorithm
• Create a map of all contig positions within each
  file
• Concatenate contigs in each file
• Run MUMmer to find MUMs
• Map back the matches to the separate contigs
• Cluster MUMs

22
PROmer

• Protein-based alignment program
• Input 2 multi-fasta files
• Technique
• Translate DNA into AA in all 6 reading frames
• Map each protein to DNA sequence
• Concatenate all potential proteins
• Run MUMmer, cluster MUMs based on DNA coordinates
• Examine a series of consecutive, consistent
  matches

23
Observation

• Align P.yeolii (5 coverage) and P.falciparum (8
  coverage) ,size 25 Mb
• PROmer time lt 1 h
• Blast time weeks
• gt70 of human chromosome 14 is duplication of
  part of chromosome 2
• Align E.coli (4.7 Mb) and V.cholerae (3 Mb) on 1
  GHz desktop computer
• MUMmer 1 74 s, 293 MB
• MUMmer 2 27 s, 100 MB

24
MUMmer 3.0

• Requires 25 less memory
• Open source
• View results with graphical UI
• Align Human vs human genome
• Computer Sun-Sparc, Solaris OS,64 GB, 950 MHz
• Size 2,839 Mbps
• Time suffix tree, 4.7 h 4 GB Memory query,
  101.5 h Total 4.5 days

25
References

• Kurtz et al. (2004) Versatile and open software
  for comparing large genomes, Genome Biology 2004,
  5R12
• Delcher et al. (2002) Fast algorithms for
  large-scale genome alignment and comparison,
  Nucleic Acids Res. 2478-2483
• Delcher et al. (1999) Alignment of whole genomes,
  Nucleic Acids Res., 27,2369-2376
• Gusfield, D. (1997) Algorithms on Strings, Trees
  and SequencesComputer Science and Computational
  Biology
• http//neo.bu.edu/kasif/be777/HarvardFeb2002.ppt
  
• http//theorie.informatik.uni-ulm.de/Lehre/SS4/CCG
  /MUMmer.ppt
• http//www.sna.csie.ndhu.edu.tw/lung/seminar/whol
  e.ppt