On restrictions of balanced 2-interval graphs - Philippe GAMBETTE

2-interval graphs. I is a realization of 2-interval graph G. a vertex a pair of intervals an edge between two vertices the pairs of intervals have a non-empty.
486KB taille 14 téléchargements 269 vues
WG'07 - Dornburg

On restrictions of balanced 2-interval graphs Philippe Gambette and Stéphane Vialette

Outline • Introduction on 2-interval graphs • Motivations for the study of this class • Balanced 2-interval graphs • Unit 2-interval graphs • Investigating unit 2-interval graph recognition

2-interval graphs 2-interval graphs are intersection graphs of pairs of intervals a vertex I

5

1

a pair of intervals 8

2

3

4

6

9 7

the pairs of intervals have a non-empty intersection

an edge between two vertices

5

1

G

8 9

4

2 3

6

7

I is a realization of 2-interval graph G.

Why consider 2-interval graphs? A 2-interval can represent : - a task split in two parts in scheduling When two tasks are scheduled in the same time, corresponding nodes are adjacent.

Why consider 2-interval graphs? A 2-interval can represent : - a task split in two parts in scheduling - similar portions of DNA in DNA comparison The aim is to find a large set of non overlapping similar portions, that is a large independent set in the 2-interval graph.

Why consider 2-interval graphs? A 2-interval can represent: - a task split in two parts in scheduling - similar portions of DNA in DNA comparison - complementary portions of RNA in RNA secondary structure prediction Primary structure: AGGUAGCCCUAGCUUAGUACUUGUCUCACUCCGCACCU

Secondary structure:

C U C A C G G C 2 A G G A U U U U C C C A G U A U A U 1 C U G G C C C AC U U C 3

RNA secondary structure prediction U

A

Helices: sets of contiguous base A pairs, appearing successive, or C U nested, in the primary structure. U C I2 I3 I1 G U C I2 C I2 G A successive nested U C U G UUCGU C Find the maximum set of disjoint G successive or nested 2-intervals: G AAGCA dynamic programming. U C UC C G C I1 A A 3 G I A helices C GU G U G G U A U C A A

RNA secondary structure prediction

Pseudo-knot: crossing base pairs. I1

I1

I2

crossed

I2

5' extremity or the RNA component of human telomerase From D.W. Staple, S.E. Butcher, Pseudoknots: RNA structures with Diverse Functions (PloS Biology 2005 3:6 p.957)

Why consider 2-interval graphs? A 2-interval can represent: - a task split in two parts in scheduling - similar portions of DNA in DNA comparison - complementary portions of RNA in RNA secondary structure prediction AGGUAGCCCUAGCUUAGUACUUGUCUCACUCCGCACCU 5

1

3

4

6 C U C A C G G C 2 A G G A U U U U C C U C A A G A U U 1 C U G G C C C AC U U C 3

8

2 9

7 5

1

8 9

4

2 3

6

7

Why consider 2-interval graphs? A 2-interval can represent:

Both intervals have same size!

- a task split in two parts in scheduling

- similar portions of DNA in DNA comparison - complementary portions of RNA in RNA secondary structure prediction AGGUAGCCCUAGCUUAGUACUUGUCUCACUCCGCACCU 5

1

3

4

6 C U C A C G G C 2 A G G A U U U U C C U C A A G A U U 1 C U G G C C C AC U U C 3

8

2 9

7 5

1

8 9

4

2 3

6

7

Restrictions of 2-interval graphs We introduce restrictions on 2-intervals: - both intervals of a 2-interval have same size: balanced 2-interval graphs - all intervals have the same length: unit 2-interval graphs - all intervals are open, have integer coordinates, and length x: (x,x)-interval graphs

Inclusion of graph classes 2-inter

perfect

AT-free

K1,4-free

circle

Ko sto ch ka claw-free ,W es t, 1 99 9

circ-arc odd-anti cycle-free

co-compar

compar chordal

trapezoid

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

Following ISGCI

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Some properties of 2-interval graphs Recognition: NP-hard (West and Shmoys, 1984) Coloring: NP-hard from line graphs Maximum Independent Set: NP-hard (Bafna et al, 1996; Vialette, 2001) Maximum Clique: open, NP-complete on 3-interval graphs (Butman et al, 2007)

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

compar

claw-free chordal trapezoid circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Balanced 2-interval graphs 2-interval graphs do not all have a balanced realization. Proof: Idea: a cycle of three 2-intervals which induce a contradiction. I1 B1

I2 B2

B3

l (I 2) < l (I 1)

B4

l (I 3) < l (I 2)

l (I 3) < l (I 1)

I3 B5

B6

l (I 1) < l (I 3)

Build a graph where something of length>0 (a hole between two intervals) is present inside each box Bi.

Balanced 2-interval graphs 2-interval graphs do not all have a balanced realization. Proof: Gadget: K5,3, every 2-interval realization of K5,3 is a contiguous set of intervals (West and Shmoys, 1984)

has only « chained » realizations:

Balanced 2-interval graphs 2-interval graphs do not all have a balanced realization. Proof: Gadget: K5,3, every 2-interval realization of K5,3 is a contiguous set of intervals (West and Shmoys, 1984)

has only « chained » realizations:

Balanced 2-interval graphs 2-interval graphs do not all have a balanced realization. Proof: Example of 2-interval graph with no balanced realization:

has only unbalanced realizations: I1 I2

I3

Recognition of balanced 2-interval graphs Recognizing balanced 2-interval graphs is NP-complete. Idea of the proof: Adapt the proof by West and Shmoys using balanced gadgets. A balanced realization of K5,3:

length: 79

Recognition of balanced 2-interval graphs Recognizing balanced 2-interval graphs is NP-complete. Idea of the proof: Reduction of Hamiltonian Cycle on triangle-free 3-regular graphs, which is NP-complete (West, Shmoys, 1984).

Recognition of balanced 2-interval graphs For any 3-regular triangle-free graph G, build in polynomial time a graph G' which has a 2-interval realization (which is balanced) iff G has a Hamiltonian cycle. Idea: if G has a Hamiltonian cycle, add gadgets on G to get G' and force that any 2-interval realization of G' can be split into intervals for the Hamiltonian cycle and intervals for a perfect matching.

G

depth 2

=

U

Recognition of balanced 2-interval graphs Recognizing balanced 2-interval graphs is NP-complete. G'

v0 v1 z M(v1)

M(v0) H3

H1

H2

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

compar

claw-free chordal trapezoid circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

compar

claw-free chordal trapezoid circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Circular-arc and balanced 2-interval graphs Circular-arc graphs are balanced 2-interval graphs Proof:

Circular-arc and balanced 2-interval graphs Circular-arc graphs are balanced 2-interval graphs Proof:

Circular-arc and balanced 2-interval graphs Circular-arc graphs are balanced 2-interval graphs Proof:

Circular-arc and balanced 2-interval graphs Circular-arc graphs are balanced 2-interval graphs Proof:

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

compar

claw-free chordal trapezoid circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

unit-2-inter

compar

claw-free chordal trapezoid

(2,2)-inter

circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

Take the left-most and the one it intersects.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

Increment their length to the right and translate the ones on the right.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

Take the left-most and the one it intersects.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

Increment their length to the right and translate the ones on the right.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of inclusion: How to transform a (x,x)-realization into a (x+1,x+1)-realization? Consider each interval separately.

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Proof of strictness: Gadget: K4,4-e, every 2-interval realization of K4,4-e is a contiguous set of intervals. I1 I2 I3 I4

I5 I6 I7

I8 I5 I6

I8

K4,4-e has a (2,2)-interval realization!

I1 I2 3 4 I I

I7

(x,x)-interval graphs The class of (x,x)-interval graphs is strictly included in the class of (x+1,x+1)-interval graphs for x>1. Idea of the proof of strictness: For x=4: any 2-interval realization of G4 has two “stairways” which requires “steps” of length at least 5.

a

vl1

X1

vr1 a

X2

G4 X2

vl1

1 r

v2 v'2

v

vl4

vr4

v3 v'3

X4

v4

vr2

X3

vr3

v

vr2

v

vr3

2 l 3 l

X3

v'4

v'1 v'2 v'3 v'4

vl3 vl2

v'1

X1

b v1 v2 v3 v4

v1

X4 vl4 b

vr4

(x,x)-interval graphs {unit 2-interval graphs} = U {(x,x)-interval graphs} x>0

Proof of the inclusion: There is a linear algorithm to compute a realization of a unit interval graph where interval endpoints are rational, with denominator 2n (Corneil et al, 1995). Corollary: If recognizing (x,x)-interval graphs is polynomial for all x then recognizing unit 2-interval graphs is polynomial.

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

unit-2-inter

compar

claw-free chordal trapezoid

(2,2)-inter

circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

unit-2-inter

compar

claw-free chordal trapezoid

(2,2)-inter

circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Proper circular-arc and unit 2-interval graphs Proper circular-arc graphs are unit 2-interval graphs Proof:

Proper circular-arc and unit 2-interval graphs Proper circular-arc graphs are unit 2-interval graphs Proof:

Proper circular-arc and unit 2-interval graphs Proper circular-arc graphs are unit 2-interval graphs Proof:

Proper circular-arc and unit 2-interval graphs Proper circular-arc graphs are unit 2-interval graphs Proof:

proper = unit

Proper circular-arc and unit 2-interval graphs Proper circular-arc graphs are unit 2-interval graphs Proof:

+ disjoint intervals

Inclusion of graph classes 2-inter balanced 2-inter

perfect

AT-free

K1,4-free

circle co-compar

unit-2-inter

compar

claw-free chordal trapezoid

(2,2)-inter

circ-arc odd-anti cycle-free

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Inclusion of graph classes 2-inter balanced 2-inter

Quasi-line graphs: every vertex is AT-free bisimplicial (its neighborhood can be partitioned into 2 cliques). K1,4-free

perfect

circle co-compar

unit-2-inter

claw-free

compar chordal

trapezoid circ-arc

(2,2)-inter

odd-anti cycle-free

quasi-line

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Inclusion of graph classes 2-inter balanced 2-inter

Quasi-line graphs: every vertex is AT-free bisimplicial (its neighborhood can be partitioned into 2 cliques). K1,4-free

perfect

circle co-compar

unit-2-inter

claw-free

compar chordal

trapezoid circ-arc

(2,2)-inter

odd-anti cycle-free

quasi-line

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Inclusion of graph classes 2-inter balanced 2-inter all-4-simp

K1,5-free

perfect

AT-free

K1,4-free

circle co-compar

unit-2-inter

claw-free

compar chordal

trapezoid circ-arc

(2,2)-inter

odd-anti cycle-free

quasi-line

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Recognition of all-k-simplicial graphs A graph is all-k-simplicial if the neighborhood of a vertex can be partitioned in at most k cliques. Recognizing all-k-simplicial graphs is NP-complete for k>2. Proof: Reduction from k-colorability. G k-colorable iff G' all-k-simplicial,

G

G'

where G' is the complement graph of G + 1 universal vertex

Inclusion of graph classes 2-inter balanced 2-inter all-4-simp

K1,5-free

perfect

AT-free

K1,4-free

circle co-compar

unit-2-inter

claw-free

compar chordal

trapezoid circ-arc

(2,2)-inter

odd-anti cycle-free

quasi-line

outerplanar bipartite

proper circ-arc = circ. interval line unit circ-arc middle

interval

unit = proper interval

co-comp int. dim 2 height 1

permutation

trees

Unit 2-interval graph recognition Complexity still open. Algorithm and characterization for bipartite graphs: A bipartite graph is a unit 2-interval graph (and a (2,2)-interval graph) iff it has maximum degree 4 and is not 4-regular. Linear algorithm based on finding paths in the graph and orienting and joining them.

Perspectives

Recognition of unit 2-interval graphs and (x,x)-interval graphs remains open. The maximum clique problem is still open on 2-interval graphs and restrictions.

Perspectives

Recognition of unit 2-interval graphs and (x,x)-interval graphs remains open. The maximum clique problem is still open on 2-interval graphs and restrictions.

Guten Appetit!