31
Sorting Cancer Karyotypes Sorting Cancer Karyotypes by Elementary Operations by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

  • View
    215

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

Sorting Cancer Karyotypes Sorting Cancer Karyotypes by Elementary Operationsby Elementary Operations

Michal Ozery-Flato and Ron ShamirSchool of Computer Science,

Tel Aviv University

Page 2: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

Outline

• Introduction

• Modeling the evolution of cancer karyotypes

• The karyotype sorting problem

• Combinatorial Analysis

• Results

2

Page 3: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

3

Introduction

Page 4: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

4

http://www.ncbi.nlm.nih.gov/sky/skyweb.cgi

Normal female

karyotype

Page 5: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

5

The "Philadelphia chromosome"

Page 6: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

6

http://www.ncbi.nlm.nih.gov/sky/skyweb.cgi

Breast cancer

karytype (MCF-7)

Page 7: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

7

Chromosomal Instability

• A phenotype of most cancer cells.– Losses or gains of chromosomes result from

errors during mitosis

– Chromosome rearrangements are associated with "double strand breaks"

multi-polar mitoses

Page 8: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

8

Double Strand Breaks

• Constitute the most dangerous type of DNA damage– A successful repair ligates two matching

broken ends– Mis-repair can result in rearrangements (e.g.

translocations) or deletions

M.C. Escher, 1953

Double strand break

Page 9: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

The Challenge

Analyze the evolution of aberration events in cancer karyotypes

9

Page 10: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

10

The Mitelman Database of Chromosome Aberrations in Cancer• Over 55,000 cancer karyotypes, culled from

over 8000 scientific publications

• Can be parsed automatically (CyDAS parser www.cydas.org)

• The largest current data resource on cancer genomes' organization

Page 11: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

11

Modeling the Evolution of Cancer Karyotypes

Page 12: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

12

The Normal Karyotype • Band = basic unit observable in karyotype. A

unique region in the genome, identified by integer

• Normal Chromosome = interval of bands– Two normal chromosomes are either disjoint or

equivalent

• Normal karyotype = a collection of normal chromosomes– Usually contains two copies of each chromosome

(with the possible exception of the sex chromosomes)

Page 13: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

13

The Cancer karyotype

• Fragment = a sub-interval (>1 bands)

of a normal chromosome

• Chromosome = – One fragment, or a concatenation of several

fragments – Orientation-less: [1,4]::[37,40] [40,37]::[4,1]

• Cancer karyotype = a collection of chromosomes

concatenation (breakpoint)

Page 14: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

14

Elementary Operations

Breakage

Fusion

du

plicatio

n

deletion

These operations can generate all known chromosomal aberrations!

Page 15: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

15

The Karyotype Sorting Problem

Page 16: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

16

The Karyotype Sorting (KS) Problem• Find a shortest sequence of elementary

operations that transforms the normal karyotype into given cancer karyotype

• Find the elementary distance = #operations in such a solution to KS.

???

Page 17: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

17

The Karyotype Sorting (KS) Problem(inverse formulation)

• Find a shortest sequence of inverse elementary operations that transforms the given cancer karyotype into the normal karyotype

???

Page 18: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

18

Inverse Elementary Operations

Breakage

Fusion

du

plicatio

n

deletion

c-d

eletion

addition

Page 19: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

19

Assumptions

• ~95% of the karyotypes in the Mitelman Database have no recurrent breakpoints

• Assumptions:– The cancer karyotype contains no recurrent

breakpoints– Every added chromosome contains no

breakpoints

[20,39]::[12,1]Breakpoint ID={390,120}

Page 20: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

20

The Reduced Karyotype Sorting (RKS) Problem

• Assumptions reduced problem:– No breakpoints in the cancer karyotype

(i.e every chromosome is an interval)

– No breakpoints created by fusions / additions All the normal chromosomes are identical

1 2 3 4 5 6 7 8 9 10 110 1 2 3 4 5 6 7 8 9 10 110

The normal karyotype The cancer karyotype

breakage, fusion, c-deletion, addition

identical chromosomes

Page 21: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

21

Combinatorial Analysis

(RKS Problem)

Page 22: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

22

Extending the karyotypes

1 2 3 4 5 6 7 8 9 10 110

The normal karyotype

1 2 3 4 5 6 7 8 9 10 110

The cancer karyotype

Page 23: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

23

Parameter 1: f = #disjoint pairs of complementing interval ends

• Observation: f = -1 for fusion; f = 1 for breakage f {0,-1,-2} for c-deletion f {0,1,2} for addition

f =5

1 2 3 4 5 6 7 8 9 10 110

Page 24: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

24

The histogram

• Parameter 2: w = #bricks

• Observations:

– w is even w = 0 for breakage / fusion w {0,2} for addition / c-deletion

1 2 3 4 5 6 7 8 9 10 110

The cancer karyotype

1 2 3 4 5 6 7 8 9 10 110

The histogram

A wall with 2 bricks

A brick

Page 25: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

25

Simple Bricks• A brick is simple if

– no lower brick (in the same wall), and– no complementing interval ends

• Parameter 3: s = #simple bricks• Observation:

s {0,-1} for breakage s =0 for constrained-deletion– |s| 2 for addition

Simple bricks

1 2 3 4 5 6 7 8 9 10110

Page 26: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

26

The Weighted Bipartite Graph of Bricks

• Parameter 4: m = the minimum weight of a perfect matching

weight v-,v+:simple v-,v+:non-simple otherwise

v- < v+ 2 0 1

v+ < v- 0 2 1

1 2 3 4 5 6 7 8 9 10110

Positive bricks

Page 27: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

Results

Page 28: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

28

Main Theorem

• The elementary-distance, d, satisfies:

w/2+f+s+m-2N d 3w/2+f+s+m-2N

N = #intervals in the normal karyotype

Page 29: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

29

Results (2)

• Used the main theorem to devise a polynomial-time 3-Approximation algorithm

• Combined with a greedy heuristic on real data (95% of Mitelman DB) optimal solutions computed for 100% of karyotypes– 99.99% cases : lower bound is achieved

(hence solution is optimal)– 30 cases: lower-bound+2 but actually optimal

(manual verification)

Page 30: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

30

Summary

• A new framework for analyzing chromosomal aberrations in cancer

• A 3-approximation algorithm when there are no recurrent breakpoints – 100% success on 57,252 karyotypes (with no

recurrent breakpoints) from the Mitelman DB.

• Future work: handle recurrent breakpoints– Analyze the remaining 5% of the karyotypes in

the Mitelman DB.

Page 31: Sorting Cancer Karyotypes by Elementary Operations Michal Ozery-Flato and Ron Shamir School of Computer Science, Tel Aviv University

31

Thank for your attention.

Questions?