[PPT] - A Method for Aligning RNA Secondary Structures Jason T. L. Wang PowerPoint Presentation

SLIDE 1

1

A Method for Aligning RNA Secondary Structures

Jason T. L. Wang New Jersey Institute of Technology

J Liu, JTL Wang, J Hu and B Tian, BMC Bioinformatics, 2005

SLIDE 2

2

Outline

Introduction
Structural alignment of RNA (preliminaries,

RSmatch algorithm, software)

Experiments (RNA motif detection)
Multiple structural alignment (RMulti)
Combining RSmatch with RNAView
Conclusion and future work

SLIDE 3

3

Molecule building blocks

Protein building blocks:

– 20 types of amino acid

RNA building blocks:

– Purine: Adenine, Guanine – Pyrimidine: Cytosine, Uracil

SLIDE 4

4

RNA structure elements

RNA sequence folds to form

secondary/tertiary structure

Majority of base connections

involve two bases

– Watson-Crick: AU or CG – Non-canonical: UG or AG

Basic structure elements of RNA

SLIDE 5

5

Definition of structural components

Given an RNA sequence:

– 5’ 3’: r1r2r3…rn

Two types of structural

components[1]:

– Single bases (blue) – Bonded base pairs (red)

A U C G G G A U C G C G G A U A U G A G G C G C A U A G C G G U 5’ 3’ [1] Zuker, M. (1989) Science

SLIDE 6

6

Secondary structure constraint (1)

No common base

can be shared by any two pairs[2].

– Bad: “G” is shared by two pairs: A-G and G-C

(a) GOOD (b) BAD A U C G G G A U C G C G G A U A U G A GG C G C A U A GC G G U 5’ 3’ A U C G G G A U C G C G G A U A U G A GG C G C A G A GC G G U 5’ 3’ C CG AC Prohibited! [2] Hofacker, I.L. (2003) NAR

SLIDE 7

7

A hairpin element must

have at least 3 bases on the loop part [3].

– Bad: only two bases (A and U) present in the loop

(a) GOOD (b) BAD A U C G G G A U C G C G G A U A U G A GG C G C A U A GC G G U 5’ 3’ A U A U C G C G G A U A U G A GG C G C A U A GC G G U 5’ 3’

Secondary structure constraint (2)

hairpin Prohibited! [3] Zuker, M. (1991) NAR

SLIDE 8

8

Pseudoknots are not included[4]

(a) BAD (b) GOOD (nested structure) (c) GOOD (branching)

Secondary structure constraint (3)

A U C G G G A U C G C G G A U A U G A GG C G C A U A GC G G U 5’ 3’ A U C G G G A U C G C G G A U AU G A G G C G C A U A G C G G U 5’ 3’ GG U A U C G G G A U C G C G G A U A U G A G G C G C A U A G C G G U 5’ 3’ A A AGG C Prohibited! [4] Mathews, D.H. (1999) JMB

SLIDE 9

9

RNA secondary structure representation schemes

a. Bond annotation[5] b. Arc representation[6] c. Tree representation[7] d. Nested parenthesis representation[8]

[5] Shapiro, B. (1990) CABIOS [6] Zhang, K. (1999) CPM [7] Ma, B. (2002) TCS [8] Hofacker, I.L. (2002) JMB

SLIDE 10

10

Outline

Introduction
Structural alignment of RNA (preliminaries,

RSmatch algorithm, software)

Experiments (RNA motif detection)
Multiple structural alignment (RMulti)
Combining RSmatch with RNAView
Conclusion and future work

SLIDE 11

11

Extended circle model

Circle model[9] :

circle 0: G, C, A, G, A, A
circle 1: A, A, U, G
circle 7: C, C, G, C, G
circle 8: G, U, A, U, U, U, C

Sequential order between components: G > C > A-U > U > C-G > A-G

G U C G A A A U U A A U G GA U C G C G C G C G C U A U U U A A C G 5’ 3’ A G

circle 0 circle 1 circle 2 circle 3 circle 4 circle 5 circle 6 circle 7 circle 8

[9] Liu, J. (2005) BMC Bioinformatics

SLIDE 12

12

Hierarchical organization

circles are organized in a tree-like hierarchy

G U C G A A A U U A A U G GA U C G C G C G C G C U A U U U A A C G 5’ 3’ A G

circle 0 circle 1 circle 2 circle 3 circle 4 circle 5 circle 6 circle 7 circle 8

SLIDE 13

13

Hierarchical relationship between two structural components

(1) the same circle: e.g. each pair from G, C, G, A-U, G-C, G, A-U (2) descendant/ancestor circles: e.g. pair (G, A-U) (3) cousin circles: e.g. pairs (U, C), (A-U, G-C) and (U, G-C)

(1) (2) (3) GU CG A A A U U A A U G G A U C G C G C G C G C UA U U U A A C G 5’ 3’ A G GU CG A A A U U A A U G G A U C G C G C G C G C UA U U U A A C G 5’ 3’ A G circle GU CG A A A U U A A U G G A U C G C G C G C G C UA U U U A A C G 5’ 3’ A G

SLIDE 14

14

Partial structure induced by a structural component

GU CG A A A U U A A U G G A U C G C G C G C G C UA U U U A A C G 5’ 3’ A G GU CG A A A U U A A U G G C G C G C G C G C UA U U U 5’ 3’ GU CG A A A U U A A U G G C G C C G C G C UA U U U 5’ 3’

10 30

parent structure child structure

SLIDE 15

15

Structural alignment rules (1)

A1 precedes A2 iff B1 precedes B2 where A1 , A2 , B1 ,B2

are structural components.

SLIDE 16

16

Structural alignment rules (2)

RNA 1 RNA 2 (a) (b) (c)

(a) Same loop relationship preserved: A1 is in the same loop as A2 iff B1 is in the same loop as B2 (b) Ancestor/descendant relationship preserved: A1 is ancestor of A2 iff B1 is ancestor of B2 (c) Cousin relationship preserved: A1 is cousin of A2 iff B1 is cousin of B2

SLIDE 17

17

Example alignment

All structural alignment

rules must be satisfied for a valid alignment

In addition, a single base

can not be aligned with a base pair

GU CG A A A U U A G C A G C G C G C G C G C G C UA U U U A A U G 3’ A U 5’ G C U CU A U U A U A A GC G G C G A U G C U A U U U A A U 3’ U A GC 5’ First RNA Second RNA Alignment Result

..((...(((......)))((.(.....))).)).. GUACGCAGUAAGUCGAUACGCCGUAUUUCGCGGUAA ..((..((......))(((.......))).)).. GUUCGAUUUCUCUAAAGAGUAGCUUUCUCGGAAA

..((...(((......)))((.(.. ...))).)).. GUACGCAGUAAGUCGAUACGCCGUA—-UUUCGCGGUAA || || | || | | | ||| |||| ||| || GUUCGA-UU-UCUCUA-AAGA-GUAGCUUUCUCGGAAA ..((.. (( ...... ))(( (.......))).))..

SLIDE 18

18

Dynamic programming algorithm:

verview

A UA C A UG U U 5’ 3’ A UC U CA U A U GA G C U A G G 5’ 3’

First structure Second structure DP scoring table

A-U A U C A U G U A U U C A U C A G G U A-U A G C-G The best alignment between partial structures

f U and A-U

SLIDE 19

19

Case 1

5’ 3’ 5’ 3’

SLIDE 20

20

Case 2

5’ 3’ 3’ 5’

SLIDE 21

21

Case 3

5’ 5’ 3’ 3’

SLIDE 22

22

Case 4.1

3’ 3’ 5’ 5’

SLIDE 23

23

Case 4.2

5’ 5’ 3’ 3’

SLIDE 24

24

Example of matching score function

Score function of matching two equal-length

structural components: i.e.

     = = =

therwise

, and pairs base are and both , 2 and bases single are and both , 1 ) , (

b a b a b a b a b a

C C C C if C C C C if C C g

Gap penalty equals 0
Extending g to the whole set of matched component pairs,
ur goal is to maximize f(R1, R2)

) , ( ) , (

2 1

i i

b a i

C C g R R f

∑

=

SLIDE 25

25

Cell type 1 : single base vs. single base

A U A C A U G U U 5’ 3’ C A UC U CA U A U GA G C U A G G 5’ 3’

..(.....).

-AUACAUGUUC

UCAUACAGGUUA ....(.....).

(A)

A U A C A U G U U 5’ 3’ C 5’

(B)

A UC U CA U A U GA G C U A G G 3’

..(.....) .

-AUACAUGUU-C

UCAUACAGGUUA- ....(.....).

A U A C A U G U U 5’ 3’ C A UC U CA U A U GA G C U A G G 5’ 3’

(C)

..(.....).

-AUACAUGUUC-

UCAUACAGGUU-A ....(.....) .

A U A C A U G U U 5’ 3’ C A UC U CA U A U GA G C U A G G 5’ 3’

AUACAUGUUC ..(.....). UCAUACAGGUUA ....(.....).

?

SLIDE 26

26

Cell type 2: base pair vs. single base

A UC U CA U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C A UC U CA U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C A U A C A U G U U 5’ 3’ C A UC U CA U A U GA G C U A G G 5’ 3’

first score second score ? ? ?

SLIDE 27

27

Cell type 2: base pair vs. single base (first score)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

ACAUGUU (.....) UCAUACAGGUUA ....(.....).

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

(.....)

---ACAUGUU-

UCAUACAGGUUA ....(.....).

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

( ..... ) A-----CAUGU--U

UCAUACAGGUUA

....(.....).

SLIDE 28

28

Cell type 2: base pair vs. single base (second score)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

AUACAUGUU ..(.....) UCAUACAGGUUA ....(.....).

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

..(.....)

-AUACAUGUU-

UCAUACAGGUUA ....(.....).

(A)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

.. (.....) AU----ACAUGUU-

-UCAUACAGGUUA

....(.....).

(B)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

(C)

.. (.....)

-AU--------ACAUGUU

UCAUACAGGUUA------- ....(.....).

SLIDE 29

29

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

AUACAUGUU ..(.....) UCAUACAGGUU ....(.....)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

? (A) (B) (C) (b1) (b2)

Cell type 3: base pair vs. base pair

?

SLIDE 30

30

Cell type 3: base pair vs. base pair (first score)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

(.....) ACAUGUU (.....) ACAGGUU

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

(.....) ACAUGUU ACAGGUU (.....)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

( ..... ) A-CAUGU-U

ACAGGUU-

(.....)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

(.....)

ACAUGUU-

A CAGGU U ( ..... )

(A) (B) (C)

SLIDE 31

31

Cell type 3: base pair vs. base pair (2nd & 3rd score)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

( ..... ) A-----CAUGU-U

UCAUACAGGUU-

....(.....)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

(.....)

---ACAUGUU

UCAUACAGGUU ....(.....)

A UC U CA U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

(.....) ACAUGUU ....(.....) UCAUACAGGUU (.....) ACAUGUU-------

-UCAU-ACAGGUU

.... (.....)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C A UC U CA U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

..(.....) AUACAUGUU (.....) ACAGGUU

SLIDE 32

32

Cell type 3: base pair vs. base pair (final score)

A UC U C A U A U G A G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

?

..(.....) AUACAUGUU ....(.....) UCAUACAGGUU

A UC U C A U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

..(.....)

-AUACAUGUU

UCAUACAGGUU ....(.....)

A UC U C A U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C A UC U C A U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C A UC U C A U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C A UC U C A U A U GA G C U A G G 5’ 3’ A U A C A U G U U 5’ 3’ C

.. (.....) AU----ACAUGUU

-UCAUACAGGUU

....(.....)

(A) (B) (C)

..(.....)

---AUACAUGUU

UCAU--ACAGGUU .... (.....)

(D)

.. (.....)

-AU-------ACAUGUU

UCAUACAGGUU------- ....(.....) ..( .....) AUA-CAUGUU-------

---UCAU---ACAGGUU

.... (.....)

SLIDE 33

33

Analysis of algorithm

Time and space complexity

– Each score is calculated only once. – Time is bounded by the number of score calculations needed to fill up the table. – Each base pair will contribute to two or four score calculations. – Single bases: Ns; base pairs: Np – Total number of score calculations: Ns

2+4NsNp+4Np 2 =O(N2)

Ns

2 score calculations are contributed by two single bases

4NsNp score calculations are contributed by one single base and one base

pair

4Np

2 score calculations are contributed by two base pairs

SLIDE 34

34

Software RSmatch

http://aria.njit.edu/rnacenter/RSmatch/

SLIDE 35

35

Outline

Introduction
Structural alignment of RNA (preliminaries,

RSmatch algorithm, software)

Experiments (RNA motif detection)
Multiple structural alignment (RMulti)
Combining RSmatch with RNAView
Conclusion and future work

SLIDE 36

36

Motif example: detection/instantiation

Motif structure is

known

IUB ambiguity

symbols:

– N: A U C G – W: A U – H: not G

SLIDE 37

37

Gap Penalty Example

motif structure subject structure

SLIDE 38

38

Position independent scoring matrices

Two scoring matrices
Gap penalty: -3 for

each single base, -6 for each base pair, involved in the gap

SLIDE 39

39

Motifs used in the experiments

HSL3 has a typical stem loop

structure with two flanking tails

IRE has specific stem-loop

structure for gene regulation related to cell iron metabolism

Wildcard “n” is allowed to

match with 0 or 1 nucleotide

IUB code:

– M: A, T/U; – Y: C, T/U; – H: not G; – R: A, G; – W: A, T;

(a) HSL3 (b) IRE

SLIDE 40

40

Experiments

Performance measurements: sensitivity (recall)

and specificity (precision)

19,986 human RefSeq mRNA sequences were
btained from NCBI; 39,972 UTR regions were

extracted

Each UTR sequence was chopped and folded into

secondary structures using Vienna RNA package, yielding ~575,000 structures

Compare RSmatch with PatSearch[10]

[10] Pesole G. (2000) Bioinformatics

SLIDE 41

41

Chop and fold UTR sequences

50 100 150 200

ORF

AAAAAAAAAAAAA

3’UTR

50 100 150 200

ORF

5’UTR

ORF: Open Reading Frame

SLIDE 42

42

Detecting HSL3 motif

PatSearch: specificity (98.2%), sensitivity (87.1%).
Several histone genes (i.e. NM_003542, NM_003548) were found by

RSmatch, but not by PatSearch.

SLIDE 43

43

Detecting IRE motif

Use PatSearch to search 39,972 UTR sequences for

IRE motif and get 27 hit structures belonging to 18 UTR sequences

The 18 UTR sequences were chopped and folded

into 1,196 structures

Compare RSmatch, Rsearch[11] and stemloc[12].
A well-known IRE-containing structure

(NM_000032) was used as the query (it does not have wildcard or ambiguity symbols since Rsearch and stemloc cannot handle them)

[11] Klein, R.J. (2003) BMC Bioinformatics [12] Holms, I. (2002) PSB

SLIDE 44

44

Experimental results for IRE motif

SLIDE 45

45

Dealing with complex structures

SLIDE 46

46

Outline

Introduction
Structural alignment of RNA (preliminaries,

RSmatch algorithm, software)

Experiments (RNA motif detection)
Multiple structural alignment (RMulti)
Combining RSmatch with RNAView
Conclusion and future work

SLIDE 47

47

Extension to multiple structural alignment

search small database seed alignment profile expand seed alignment

score (best alignment) < δ OR non-expandable

expand best alignment pairwise match

NO YES

SLIDE 48

48

Example

expand expand

SLIDE 49

49

RMulti Webserver

http://aria.njit.edu/rnacenter/multi.html

SLIDE 50

50

Outline

Introduction
Structural alignment of RNA (preliminaries,

RSmatch algorithm, software)

Experiments (RNA motif detection)
Multiple structural alignment (Rmulti)
Combining RSmatch with RNAView
Conclusion and future work

SLIDE 51

51

SLIDE 52

52

SLIDE 53

53

Outline

Introduction
Structural alignment of RNA (preliminaries,

RSmatch algorithm, software)

Experiments (RNA motif detection)
Multiple structural alignment (RMulti)
Combining RSmatch with RNAView
Conclusion and future work

SLIDE 54

54

Conclusion

An efficient algorithm RSmatch to align

and analyze RNA secondary structures

A multiple structural alignment tool RMulti
A visualization tool combining RSmatch

with RNAView

SLIDE 55

55

Future Work

Extending RSmatch to handle pseudoknots
Large-scale genome-wide motif mining
Indexing very large RNA structure databases
Improved multiple structural alignment of RNA

sequences

RNA classification and clustering
RNA-RNA interactions and protein-RNA

interactions

SLIDE 56

56