Algorithms for Searching RNA Motifs in Genomic DNA Sequences

1
Algorithms for Searching RNA Motifs in Genomic
DNA Sequences

• JCIS 2005 PRESENTATION
• Jingping Liu
• Bin Ma, Kaizhong Zhang

University of Western Ontario, Computer Science
Department July, 2005
2
Outline

• Motivations
• RNA Structures
• Definitions Notations
• Tree Representation Algorithms
• Experimental Results
• Conclusions Further Work

3
Motivations

• Bio-molecule structures are invaluable in
  endeavors such as creating new drugs and
  understanding genetic diseases.
• Computational algorithms for genome analysis are
  desirable. Fichants tRNAscan, Pavesis
  EufindtRNA are popular tools - - Only for
  computing tRNA!
• Recently, Zhangs HomoStRscan for detecting RNA
  has been introduced. Designing algorithms
  functioning as a filter of Zhangs approach is
  one of our primary motivations.

4
RNA Structures

• RNA Primary structure is a sequence of bases
  (i.e., ? A, C, G, U).
• RNA Secondary structure is a set of base pairs
  (i.e., A-U, C-G, G-U, and vice versa). Assume
  (1) no base takes part in more than one base
  pair (2) Base pairs never cross.

A
G
C
U
G
C
U
G
U
A
C
G
U
A
A
A
(a)
(b)
5
Example of RNA Secondary Structure
Loop-Segment
5'

• The paired region is called stem
• The unpaired region is called loop-segment.
• The stem-loop consists of a stem surmounted by a
  loop-segment (a.k.a. hairpin loop).

3'
Stem
Stem-Loop
6
Problem Statement
RNA Structure
In a given genomic sequence, efficiently
determine candidate segments that can
potentially form RNA secondary structures
similar to a given RNA secondary structure.
Genomic Sequence


A
G
C
U
G
C
U
G
U
A
C
G
U
A
A
A
p
q
Candidate Segment
One sequence may form many different secondary
structures!
7
Definitions Notations

• In a stem, the number of base pairs is called
  stem size, denoted by SM.
• In a loop-segment, the number of unpaired bases
  is called loop-segment size, denoted by D.
• Hairpin size the number of unpaired bases in a
  hairpin loop, denoted by H.

U
G
U
C
Hairpin Loop
U
U
C
G
U
A
C
G
(H 4)
Loop-segment
A
U
(D 6)
U
A
Stem
G
C
(SM 4)
8
Definitions Notations (contd)

• Similar Stem ?SM/2? SMx (SM ?SM/2?)
• Similar Loop-segment ?D/2? Dx (D ?D/2?)

(1) Similar Stems
C
G
C
G
U
A
C
G
C
G
A
U
A
U
G
C
A
U
A
U
U
A
U
A
U
A
U
A
2
3
G
C
G
C
G
C
G
C
A
U
SM4
5
U
A
6
(2) Similar Loop-segments
U
U
C
G
U
A
U
C
G
U
A
U
U
C
G
U
A
C
U
A
6
2
D4
5
3
9
Tree Representation

• Leaf node -- represents stem-loop, and contains
  necessary values. e.g. Hmin, Hmax of hairpin
  loop.
• Internal node (including root) -- represents
  stem, and contains necessary values. e.g. SMmin,
  SMmax of stem, and Dmin, Dmax of loop segments.

10
Example of Tree Representation
3
A
C
C
5
A
(1) RNA
1
72
G
C
G
C
C
G
(2) Tree
U
A
G
C
n4
G
C
65
G
C
G
10
G
A
U
U
A
C
G
C
C
C
A
A
A
U
G
C
C
49
U
G
U
G
A
G
A
n3
n1
n2
U
28
44
A
A
C
G
G
C
A
U
A
C
C
26
G
A
U
U
U
U

C
G
U
A
C
G
A
U
C
C
U
U
A
G
A
11
Bottom-Up Approach

• Leaf node (stem-loops) using SEARCH
• 2. Internal node (stem) with degree d
• Integrate step
• Extend step
• Repeat integrate and extend steps until root.
• Reduce the number of candidates by using
• additional biological features.

n4
RNA Tree
n3
n1
n2
12
SEARCH Algorithm

• SEARCH Based on corresponding min and max values
  of sizes, search for one potential stem-loop/stem
  in the given genomic sequence S.
• Given an index-pair (i, j), let i go left and j
  go right to search each possible stem-loop/stem
  in S by looking for consecutive base pairs
  (i.e., A-U, C-G and G-U, and vice versa) until no
  further such base pair.

SMmax
i
j
13
Integrate and Extend Steps

• INTEGRATE (left-to-right) If loop-segment size
  D of two located substructures satisfies Dmin ? D
  ? Dmax, then integrate them to form a more
  complicated substructure (SP, EP).
• EXTEND Use SEARCH algorithm, and start at (SP,
  EP) to compute the extended stem, such that each
  size satisfies corresponding min and max values.

1
(2) Tree
(1) RNA
D12
integrate steps step 1, and 2.
D41
SP
2
n4
EP
D23
4
D34
3
n3
n1
n2
14
mSearch Experimental Results
tRNA
5S rRNA
Sequence
NCBI
mSearch
Sequence
NCBI
mSearch
Mycoplasma Genitalium ( 0.4M)
Staphylococcus MW2 (partial 0.4M)
992 ( 5s)
36
1986 ( 5s)
3
Helicobacter Pylori ( 1.7M)
Escherichia Coli K12 (partial 0.4M)
1349 ( 5s)
4
36
1640 ( 20s)
Experimental results indicate that our mSearch
tool can locate all true tRNAs and 5S rRNAs in
the experimented sequences - - zero false
negative !
15
Conclusions Further Work

• Theoretical and Practical aspects
• Develop efficient pattern matching algorithms
• Search RNA motifs in genomic sequences.
• Our algorithms are suitable for searching any
  type of RNA.
• Currently, a web-based mSearch tool is under
  construction.
• Add biological constrains to further reduce the
  number of candidates.

16
References

• D. Gusfield. Algorithms on strings, trees, and
  sequences. Computer Science and Computational
  Biology. Cambridge University Press, 143-148
  1997.
• G. Mauri, and G. Pavesi. Pattern Discovery in RNA
  Secondary Structures Using Affix Trees. CMP 278 -
  294, 2003.
• K. Zhang, S. Le, and J. Maizel. An Algorithm for
  Detecting homologues of Known Structured RNAs in
  Genomes. IEEE Computational Systems, CBS 2004.
• K. Zhang, B. MA, and L.Wang. Computing
  Similarity between RNA Structures. Theor. Comput.
  Sci. 276(1-2) 111-132 2002.
• M. Zuker. The Use of Dynamic Programming
  Algorithms in RNA Secondary Structure Prediction,
  in Mathematical Methods for DNA Sequences. CRC
  Press, Inc.,Boca Raton, 1989, 159.

17
QUESTION ?
THANK YOU