Sequence alignment with rearrangements

1
Sequence alignment with rearrangements

• Pankaj Kumar
• Jitesh Kumar
• Shouvik Som

2
Alignments

• Global
• One string is transformed into the other.
• Less prone to demonstrating false homology.
• Each letter of one sequence is constrained to
  being aligned to only one letter of other.
• Needleman Wunsch (1970), LAGAN(2003)

• Local
• All locations of similarity between 2 strings r
  returned.
• Higher false positive rates cant take overall
  conservation maps.
• CHAOS(2002),BLASTZ(2003).

3
Rearrangement Events What are they ?

• DNA mutates by various rearrangement events,
  during evolution. They are -
• Translocation a subsegment removed inserted in
  a different location but same orientation.
• Inversion subseg removed from seq then
  inserted in same location but in different
  orientation.
• Duplication a copy of subseg inserted into the
  seq, the original subseg is unchanged.
• Or, a combination of above.

4
Glocal Aligner its need?

• Neither Global, nor Local aligners handle
  rearrangements events satisfactorily.
• Global Aligners dont handle these events at all,
  as they should be monotonically increasing in
  both sequences.
• Local Aligners, dont suggest how sequences could
  have evolved from a common ancestor.
• So, a Hybrid.., a glocal aligner capable of
  quickly aligning long genomic sequences has to be
  used.

5
Glocal Alignment

• A glocal alignment between two sequences is a
  series of operations that transform one sequence
  into the other.
• The necessary set of operations are includes
  insertions, deletions, point mutations,
  inversions, translocations, and duplications.
• Each operation, incurs a penalty, and total edit
  distance, is the sum of penalties.
• One such Glocal Alignment Algorithm is SLAGAN.
• (Michael Brudno1,, Sanket Malde1,, Alexander
  Poliakov 2,
• Chuong B. Do1, Olivier Couronne2, Inna Dubchak2
  and Serafim Batzoglou 1,).

6
The Shuffle-LAGAN algorithm

• Consists of three distinct stages.
• Generation of local alignments using CHAOS tool.
• Building the 1-monotonic conservation map.
• Aligning consistent subsegments using LAGAN
  Global Aligner.

7
Generation of Local Alignments

• SLAGAN uses CHAOS (Brudno and Morgenstern, 2002
  Brudno et al., 2003) a method that finds small
  matching words with degeneracy, and chains them
  into local alignments.

8
Working with Chaos

• ./chaos sample.fasta dbase.fasta -co 4 -b -wl 5
• co cutoff score b both strands wl
  word length
• sample.fasta
• gtsample1
• AAATGTCC
• dbase.fasta
• gtsample2
• GGCATGTCCAGAAAATCCAAGTGCCTCTTCCTCTTGATCTTCTCCAACGA
  TGTCCAGA
• AAATCCAAGTGCCTCATTCCTCTTGATCTTCTCCAGGCATGTCCAGAAAA
  TCCAAGTG
• CCTCTTCCTCTCTGATCTTCTCCTCGGTTGGTCCAGAAAATCCAAGTGCC
  TCTTCCTC
• TTGATCTTCTCCAGAAATGTCCAGAAAATCCAAGTAGCCTCTTCCTCTTG
  ATCGGCTC
• CAGAAATGTCCAGAAAAATCCAAGTGCCTCTTCCTCTTGATCGGCTCCAT
  AAATGTCC
• AGAAAATCCAACGTGCCTCTTCCTCTTGATCGGCTCCAGAAATGTCCAGA
  AATATCCA
• AGTGCCTCTTCCTCTTGATCGGCTCCTTA
• Chaos Output

9
CHAOS Output

• sample1 1 4 sample2 246 243 score 4100.000000
  (-)
• sample1 1 8 sample2 340 347 score 624.000000
  ()
• sample1 1 7 sample2 19 25 score 411.000000
  ()
• sample1 1 7 sample2 65 71 score 411.000000
  ()
• sample1 1 7 sample2 112 118 score 411.000000
  ()
• sample1 1 7 sample2 160 166 score 411.000000
  ()
• sample1 1 8 sample2 189 196 score 1369.000000
  ()
• sample1 1 8 sample2 236 243 score 1369.000000
  ()
• sample1 1 7 sample2 254 260 score 411.000000
  ()
• sample1 1 8 sample2 330 337 score 1369.000000
  ()
• sample1 1 7 sample2 348 354 score 411.000000
  ()
• sample1 1 4 sample2 59 62 score 364.000000
  ()
• sample1 1 4 sample2 106 109 score 364.000000
  ()
• sample1 1 4 sample2 154 157 score 364.000000
  ()
• sample1 2 8 sample2 3 9 score 459.000000 ()
• sample1 2 8 sample2 49 55 score 542.000000
  ()
• sample1 2 8 sample2 96 102 score 459.000000
  ()
• sample1 2 8 sample2 300 306 score 317.000000
  ()
• sample1 5 8 sample2 147 150 score 391.000000
  ()

10
Building the 1-monotonic map

• What is a 1-Monotonic Chain?
• A local alignment is represented as
• L (start1, end1, start2, end2, score, strand)
• Consider two alignments L1 L2. We call them
  1-monotonic if,
• L2.start1 gt L1.end1 .
• An ordered list of local alignments L1Lk is
  1-monotonic if for any pair of local alignments
  Li, Lj if i lt j then Li Lj are 1-monotonic
  respectively. Intuitively, a list of local
  alignments is 1-monotonic if it is strictly
  increasing in the first sequence.

11
Implementation of 1-monotonic chaining

• Parse the Chaos File
• Create a maximum score Chain using Dynamic
  Programming

12
Structure used for a local alignment

• Linked list of local alignments after parsing
  Chaos output file
• typedef struct frag
• int startx
• int endx
• int starty
• int endy
• float score
• float chainscore
• char strand
• struct frag chained_with
• struct frag chained_with_front
• struct frag forward
• struct frag backward
• Anchor
• back

frag3
frag4
frag1
frag2
13
Dynamic programming pseudo code

• for every v
• v-gtchainscore 0
• traverse every node w before v in the
  doubly linked list .
• if (chainable(w,v) cal_chainscore
  (w,v)gtv-gtchainscore)
• v-gtchainedwith w
• v-gtchainscore cal_chainscore(w,v)
  

14
Consistent Subsegments

• What are Consistent Subsegments?
• Two local alignments are consistent if -
• They are 1-monotonic.
• Both on same strand.
• L2.start2 gt L1.end2 if on ve strand L2.start2
  lt L1.end2 on ve strand.
• .

15
Implementation

• Traverse the 1-monotone chain to find consistent
  subsegments
• For every consistent subsegment creates two files
  seq1.fasta seq2.fasta, containing the
  corresponding substrings of the sequences to be
  aligned.
• LAGAN is called to align the sequences.

16
Test Case (one direct and one reversed)

• GGCATGTCCAGAAAATCCAAGTGCCTCTTCCTCTTGATCTTCTCCAACGA
  TGTCCAGA
• AAATCCAAGTGCCTCATTCCTCTTGATCTTCTCCAGGCATGTCCAGAAAA
  TCCAAGTG
• CCTCTTCCTCTCTGATCTTCTCCTCGGTTGGTCCAGAAAATCCAAGTGCC
  TCTTCCTC
• TTGATCTTCTCCAGAAATGTCCAGAAAATCCAAGTAGCCTCTTCCTCTTG
  ATCGGCTC
• CAGAAATGTCCAGAAAAATCCAAGTGCCTCTTCCTCTTGATCGGCTCCAT
  AAATGTCC
• AGAAAATCCAACGTGCCTCTTCCTCTTGATCGGCTCCAGAAATGTCCAGA
  AATATCCA
• AGTGCCTCTTCCTCTTGATCGGCTCCTTA
• GGCACTTGGATTTTCTTCTCTCTTTAAACCTCTTGATCGGCTCC

17
Chaos output

• ./chaos sample1.fasta sample2.fasta co 5 wl 10
  b
• co cutoff score wl word length b
  both strands
• sample1 28 44 sample2 217 233 score
  5315.000000 ()
• sample1 28 44 sample2 264 280 score
  5315.000000 ()
• sample1 28 44 sample2 311 327 score
  5315.000000 ()
• sample1 28 44 sample2 358 374 score
  5315.000000 ()
• sample1 28 39 sample2 29 40 score 2177.000000
  ()
• sample1 28 39 sample2 76 87 score 2382.000000
  ()
• sample1 28 39 sample2 170 181 score
  2177.000000 ()
• sample1 1 14 sample2 354 341 score
  1114.000000 (-)
• sample1 1 15 sample2 260 246 score
  2494.000000 (-)
• sample1 1 17 sample2 166 150 score
  5157.000000 (-)
• sample1 1 17 sample2 118 102 score
  5157.000000 (-)
• sample1 1 17 sample2 71 55 score 5157.000000
  (-)
• sample1 1 17 sample2 25 9 score 5157.000000
  (-)
• sample1 3 17 sample2 210 196 score
  2054.000000 (-)
• sample1 5 17 sample2 302 290 score
  3042.000000 (-)

18
Result 1

• seq1 GGCACTTGGATTTTCT-------------------------
  -------------------
• seq2 GGCACTTGGATTTTCTGGACCAACCGAGGAGAAGATCAGAG
  AGGAAGAGGCACTTGGATT
• 10 20 30 40
  50 60
• seq1 -----------------------------------------
  -------------------
• seq2 TTCTGGACATGCCTGGAGAAGATCAAGAGGAATGAGGCACT
  TGGATTTTCTGGACATCGT
• 70 80 90 100
  110 120
• seq1 -----------------------------------------
  -----------------------------------T
• seq2 TGGAGAAGATCAAGAGGAAGAGGCACTTGGATTTTCTGGAC
  ATGCC
• 130 140 150 160

19
Result 2

• seq1 CTCTCTT----------------------------------
  -------------------
• seq2 TCTTCCTCTTGATCTTCTCCAGAAATGTCCAGAAAATCCAA
  GTAGCCTCTTCCTCTTGAT
• 10 20 30 40
  50 60
• seq1 -----------------------------------------
  -------------------
• seq2 CGGCTCCAGAAATGTCCAGAAAAATCCAAGTGCCTCTTCCT
  CTTGATCGGCTCCATAAAT
• 70 80 90 100
  110 120
• seq1 -----------------------------------------
  -------------------
• seq2 GTCCAGAAAATCCAACGTGCCTCTTCCTCTTGATCGGCTCC
  AGAAATGTCCAGAAATATC
• 130 140 150 160
  170 180
• 10 20
• seq1 -------------TAAACCTCTTGATCGGCTCC---

20
Cost Function

• The gap penalty charged for all
• transitions consists of 3 parts -
• Gap Open Constant
• Gap Continue Penalty (L1.end1L1.end2)(L2.star
  t1L2.start2)constant
• Distance Alignment Penalty
• min(L1.end1L2.start1, L1.end2-L2.start2)
• back

21

• startx 1 endx 4 starty 246 endy 243
  strand -
• startx 5 endx 8 starty 147 endy 150
  strand