1. Computational Systems Biology of Cancer:

2. Professor of Computer Science, Mathematics and Cell Biology
Bud Mishra
Professor of Computer Science, Mathematics and Cell Biology
Courant Institute, NYU School of Medicine, Tata Institute of Fundamental Research, and Mt. Sinai School of Medicine

4. War on Cancer
“Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don't know we don't know.”
US Secretary of Defense, Mr. Donald Rumsfeld, Quoted completely out of context.

5. Introduction: Cancer and Genomics:
What we know & what we do not
“Cancer is a disease of the genome.”

6. Outline Genomics Genome Modification & Repair
Segmental Duplications Models

7. Genomics
Genome:
Hereditary information of an organism is encoded in its DNA and enclosed in a cell (unless it is a virus). All the information contained in the DNA of a single organism is its genome.
DNA molecule can be thought of as a very long sequence of nucleotides or bases:
S = {A, T, C, G}

8. Complementarity
DNA is a double-stranded polymer and should be thought of as a pair of sequences over S.
However, there is a relation of complementarity between the two sequences:
A , T, C , G

9. DNA Structure.
The four nitrogenous bases of DNA are arranged along the sugar- phosphate backbone in a particular order (the DNA sequence), encoding all genetic instructions for an organism.
Adenine (A) pairs with thymine (T), while cytosine (C) pairs with guanine (G).
The two DNA strands are held together by weak bonds between the bases.

10. Structure and Components
Complementary base pairs
(A-T and C-G)
Cytosine and thymine are smaller (lighter) molecules, called pyrimidines
Guanine and adenine are bigger (bulkier) molecules, called purines.
Adenine and thymine allow only for double hydrogen bonding, while cytosine and guanine allow for triple hydrogen bonding.

11. Inert & Rigid
Thus the chemical (hydrogen bonding) and the mechanical (purine to pyrimidine) constraints on the pairing lead to the complementarity and makes the double stranded DNA both chemically inert and mechanically quite rigid and stable.

12. The Central Dogma
The central dogma(due to Francis Crick in 1958) states that these information flows are all unidirectional:
“The central dogma states that once `information' has passed into protein it cannot get out again.”
DNA
RNA
Protein
Transcription
Translation

13. The Central Dogma
“…The transfer of information from nucleic acid to nucleic acid, or from nucleic acid to protein, may be possible, but transfer from protein to protein, or from protein to nucleic acid is impossible. Information means here the precise determination of sequence, either of bases in the nucleic acid or of amino acid residues in the protein.”
DNA
RNA
Protein
Transcription
Translation

14. The New Synthesis DNA RNA Protein Genome Evolution Selection
Part-lists, Annotation, Ontologies
Transcription
Translation
Genotype
Phenotype

15. Cancer Initiation and Progression
Mutations, Translocations, Amplifications, Deletions
Epigenomics (Hyper & Hypo-Methylation)
Alternate Splicing
Cancer Initiation and Progression
Proliferation, Motility, Immortality, Metastasis, Signaling

16. Multi-step Nature of Cancer:
Cancer is a stepwise process, typically requiring accumulation of mutations in a number of genes.
~6-7 independent mutations typically occur over several decades:
Conversion of proto-oncogenes to oncogenes
Inactivation of tumor suppressor gene

17. Amplifications & Deletions
Mutation in a TSG
Epigenomics
Conversion of a
Proto-Oncogene
Deletion of a TSG
Deletion of a TSG

18. P53 Gene (TSG)

19. The Cancer Genome Atlas
Obtain a comprehensive description of the genetic basis of human cancer.
Identify and characterize all the sites of genomic alteration associated at significant frequency with all major types of cancers.

20. The Cancer Genome Atlas
Increase the effectiveness of research to understand
tumor initiation and progression,
susceptibility to carcinogensis,
development of cancer therapeutics,
approaches for early detection of tumors &
the design of clinical trials.

21. Specific Goals
Identify all genomic alterations significantly associated with all major cancer types.
Such knowledge will propel work by thousands of investigators in cancer biology, epidemiology, diagnostics and therapeutics.

22. To Achieve this goal …
Create large collection of appropriate, clinically annotated samples from all major types of cancer; and
Characterize each sample in terms of:
All regions of genomic loss or amplification,
All mutations in the coding regions of all human genes,
All chromosomal rearrangements,
All regions of aberrant methylation, and
Complete gene expression profile, as well as other appropriate technologies.

23. Biomedical Rationale
Cancer is a heterogeneous collection of heterogeneous diseases.
For example, prostate cancer can be an indolent disease remaining dormant throughout life or an aggressive disease leading to death.
However, we have no clear understanding of why such tumors differ.

24. Biomedical Rationale
Cancer is fundamentally a disease of genomic alteration.
Cancer cells typically carry many genomic alterations that confer on tumors their distinctive abilities (such as the capacity to proliferate and metastasize, ignoring the normal signals that block cellular growth and migration) and liabilities (such as unique dependence on certain cellular pathways, which potentially render them sensitive to certain treatments that spare normal cells).

25. History
1960s
The genetic basis of cancer was clear from cytogenetic studies that showed consistent translocations associated with specific cancers (notably the so-called Philadelphia chromosome in chronic myelogenous leukemia).
1970s
Recognize specific cancer-causing mutations through recombinant DNA revolution of the 1970s.
The identification of the first vertebrate and human oncogenes and the first tumor suppressor genes,
These discoveries have elucidated the cellular pathways governing processes such as cell-cycle progression, cell-death control, signal transduction, cell migration, protein translation, protein degradation and transcription.
For no human cancer do we have a comprehensive understanding of the events required.

26. Scientific Foundation for a Human Cancer Genome Project
Gene resequencing.
Specific gene classes (such as kinases and phosphatases) in particular cancer types.
Epigenetic changes.
Loss of function of tumor suppressor genes by epigenetic modification of the genome — such as DNA methylation and histone modification.
Genomic loss and amplification.
Consistent association with genomic loss or amplification in many specific regions, indicating that these regions harbor key cancer associated genes
Chromosome rearrangements.
Activate kinase pathways through fusion proteins or inactivating differentiation programs through gene disruption.
Hematological malignancies: a single stereotypical translocation in some diseases (such as CML) and as many as 20 important translocations in others (such as AML).
Adult solid tumors have not been as well characterized, in part owing to technical hurdles.

27. Human Genome Structure

28. EBD J.B.S. Haldane (1932): Susumu Ohno (1970):
“A redundant duplicate of a gene may acquire divergent mutations and eventually emerge as a new gene.”
Susumu Ohno (1970):
“Natural selection merely modified, while redundancy created.”

29. Evolution by Duplication
Genome Evolution
Non-random distribution in Genomic data
Short words frequency
Protein family size
Long range orrelation
Motifs in cellular etworks
Leu3
LEU1
BAT1
ILV2
UMP
UDP
UTP
CTP
Regulatory motifs
Metabolic pathways
Interaction clusters
Functional modules
Living cells
Organism ?

30. Human Condition

31. Mer-scape…
Overlapping words of different sizes are analyzed for their frequencies along the whole human genome
Red: 24-mers,
Green: 21-mers
Blue:18 mers
Gray:15 mers
To the very left is a ubiquitous human transposon Alu. The high frequency is indicative of its repetitive nature.
To the very right is the beginning of a gene. The low frequency is indicative of its uniqueness in the whole genome.

32. Doublet Repeats
Serendipitous discovery of a new uncataloged class of short duplicate sequences; doublet repeats.
almost always < 100 bp
(Top) . The distance between the two loci of a doublet is plotted versus the chromosomal position of the first locus.
(Bottom) : Distribution of doublets (black) and segmental duplications (red) across human chromosome 2

33. Segmental Duplications
3.5% ~ 5% of the human genome is found to contain
segmental duplications, with length > 5 or 1kb, identity > 90%.
These duplications are estimated to have emerged about 40Mya under neutral assumption.
The duplications are mostly interspersed (non-tandem), and happen both inter- and intra-chromosomally.
Human
From [Bailey, et al. 2002]

34. The Model f - - f ++ f + - f - - f ++ f + -
Mutation accumulation in the duplicated sequences
f - -
f ++
f + -
insertion
deletion or mutation
Duplication by recombination between repeats
Duplication by recombination between other repeats or other mechanisms

35. The Mathematical Modelα
1-α-2β
1-α-2γ
1-α-β/2-γ γ 2β 2γ
β/2 h0 h1
h1++
h1--
h0+-
h0--
h0++
f - -
f ++
f + -
Time after duplication
0 ≤ d < ε
ε ≤ d < 2ε
(k-1)ε ≤ d < kε
h1: proportion of duplications by repeat recombination;
h1++: proportion of duplications by recombination of the specific repeat;
h1- - : proportion of duplications by recombination of other repeats;
h0: proportion of duplications by other repeat-unrelated mechanism;
h0++: proportion of h0 with common specific repeat in the flanking regions;
h0+-: proportion of h0 with no common specific repeat in the flanking regions;
h0- -: proportion of h0 with no specific repeat in the flanking regions;
α: mutation rate in duplicated sequences;
β: insertion rate of the specific repeat;
γ: mutation rate in the specific repeat;
d: divergence level of duplications;
ε: divergence interval of duplications.

36. h1Alu ≈ 0.76; h1++Alu ≈ 0.3; h1L1 ≈ 0.76; h1++ L1 ≈ 0.35.
Model Fitting
Diversity:
f - -
f ++
f + -
Alu L1
The model parameters (αAlu, βAlu, γAlu, αL1, βL1, γL1) are estimated from the reported mutation and insertion rates in the literature.
The relative strengths of the alternative hypotheses can be estimated by model fitting to the real data.
h1Alu ≈ 0.76; h1++Alu ≈ 0.3; h1L1 ≈ 0.76; h1++ L1 ≈ 0.35.

37. Polya’s Urn
Random drawn
Duplication
Put back

38. Repetitive Random Eccentric GOD
Genome Organizing Devices (GOD)
Polya’s Urn Model:
F’s: functions deciding probability distributions
F1 (decide an initial position)
F2 (decide selected length)
F3 (decide copy number)
k copies
F4 (decide insertion positions)

39. DNA polymerase stuttering (replication slippage)
Deletion:
Insertion (duplication):
normal replication
DNA polymerase
normal replication
polymerase pausing and dissociation
polymerase pausing and dissociation
3’ realignment and polymerase reloading
3’ realignment and polymerase reloading

40. Transposons ~ causes: deletions and duplications
Transposon actions in
genomic DNA:
Donor DNA
Duplication:
transposon
DNA intermediates
Target DNA
Transposase cuts in target DNA
Deletion:
transposon IS IS
Transposon looping out
Transposon inserted
Transposon deleted
DNA is repaired-resulting in a duplication of the transposon and target site

41. DNA mismatch repair mechanism ~ prevents duplications and deletions
Mismatch recognition on daughter strand
MSH6
MSH2
Degradation of the mismatched daughter strand in the a-loop
MSH6
MSH2
PMS2
MLH1
MSH2
MSH6
MLH1
PMS2
Exo
DNA a-loop formation by translocation through the proteins
Refilling the gap by DNA polymerase
Corrected daughter strand
DNA polymerase

42. Graph Model Graph description A. Deletion: With probability p0
G = {V, E}, G is a directed multi-graph;
V  {Vi, all n-mer’s, i = 1…4n}, ;
E  { (Vi , Vj ), when Vi represents the n-mer immdiately upstream of the n-mer represented by Vj in the genomic sequence};
ki = incoming (or outgoing) degree of node i (Vi) = copy number of the n-mer represented by Vi;
During the graph evolution, at each iteration, one of the following happens: deletion (with probability p0), duplication (with probability p1) or substitution (with probability q). p0 + p1 + q = 1.
For an arbitrary node Vi , the probabilities of one of the above events happens is as follows:
B. Duplication:With probability p1
C. Substitution: With probability q

43. Model Fitting Real distribution from genome analysis
Expected distribution from random sequences
Model fitting results
Model fitting results:
Initial condition.

44. Model Fitted Parameter
The substitution rate, q, increases with the sizes of mer’s.
The ratio between duplication and deletion rate, p1/p0, increases with sizes of mer’s.
The substitution rate, q, tends to decrease when the genome sizes are larger. Especially, q is much smaller in eukaryotic genomes than in prokaryotic genomes.
Mer size
6 mer
7 mer
8 mer
9 mer q
p1/p0
M. genitalium
0.0176
1.1
0.0436
0.1587
1.5
0.3222
M. pneumoniae
0.0319
0.1151
0.2309
0.4363
P. abyssi
0.0269
0.0672
1.2
0.1778
1.4
0.3897
P. horikoshii
0.0234
0.0443
0.1390
1.3
0.3456
P. furiosus
0.0213
0.0384
0.1119
0.3114
H. pylori
0.0180
0.0320
0.0925
0.2262
H. influenzae
0.0202
0.0366
0.1364
0.2802
S. tokodaii
S. subtilis
0.0187
0.0326
0.1139
0.2585
E. coli K12
0.0207
0.0334
0.0698
0.2389
S. cerevisiae
0.0113
0.0459
0.1311
C.elegans --
0.0076
0.0115
0.0275
Two independent parameters are fitted. And there are some trends:
The value of q gets bigger with the size of mer.
The value of p1/p0 also seems to get bigger with the size of mer.
The value of q seems to go smaller when genome size increases.
Q: Do the parameters have any biological meanings?
A: Yes, refer to supplementary slides – under the assumption of power—law distribution of the lengths of duplication and deletion events, the parameter can be predictive for length and frequency of duplication and deletion events in a specific genome.

45. J.B.S. Haldane
“If I were compelled to give my own appreciation of the evolutionary process…, I should say this: In the first place it is very beautiful. In that beauty, there is an element of tragedy…In an evolutionary line rising from simplicity to complexity, then often falling back to an apparently primitive condition before its end, we perceive an artistic unity …
“To me at least the beauty of evolution is far more striking than its purpose.”
J.B.S. Haldane, The Causes of Evolution

46. Human Cancer Genome

47. Cancer
Normal epithelial mucosa
Neoplastic polyp

48. A Challenge
“At present, description of a recently diagnosed tumor in terms of its underlying genetic lesions remains a distant prospect. Nonetheless, we look ahead 10 or 20 years to the time when the diagnosis of all somatically acquired lesions present in a tumor cell genome will become a routine procedure.”
Douglas Hanahan and Robert Weinberg
Cell, Vol. 100, 57-70, 7 Jan 2000

49. Karyotyping

50. CGH: Comparative Genomic Hybridization.
Equal amounts of biotin-labeled tumor DNA and digoxigenin-labeled normal reference DNA are hybridized to normal metaphase chromosomes
The tumor DNA is visualized with fluorescein and the normal DNA with rhodamine
The signal intensities of the different fluorochromes are quantitated along the single chromosomes
The over-and underrepresented DNA segments are quantified by computation of tumor/normal ratio images and average ratio profiles
Amplification
Deletion

51. CGH: Comparative Genomic Hybridization.

52. Microarray Analysis of Cancer Genome
Normal DNA
Normal LCR
Tumor DNA
Tumor LCR
Label
Hybridize
Representations are reproducible samplings of DNA populations in which the resulting DNA has a new format and reduced complexity.
We array probes derived from low complexity representations of the normal genome
We measure differences in gene copy number between samples ratiometrically
Since representations have a lower nucleotide complexity than total genomic DNA, we obtain a stronger specific hybridization signal relative to non-specific and noise

53. Copy Number Fluctuation

54. Measuring gene copy number differences between complex genomes
Hybridize to 500,000 features microarray
Genomic DNA patient samples
Statsitical Analysis  
Compare the genomes of diseased and normal samples
Error Control:
The use of representations augmenting microarrays
Representations reproducibly sample the genome thereby reducing its complexity. This increases the signal-to-noise ratio and improves sensitivity
Statistical Modeling the sources of Noise
Bayesian Analysis

55. Nattering Nabobs of Negativism
Some scientists are concerned about the cost and the possibility that a project of this scale could take money away from smaller ones…
Craig Venter, who led a private project to determine the human DNA blueprint in competition with the human genome project, said it would make more sense to look at specific families of genes known to be involved in cancer.
Lee Hood, president of the Institute for Systems Biology, has called the premise of the Cancer Genome Project “naïve,” suggesting that signal-to-noise issues its researchers are likely to encounter will be “absolutely enormous.”

56. Challenges ArrayCGH data is noisy!
Better computational biology algorithms
Better statistical modeling…In some technologies, the noise is systematic (e.g., affy SNP-chips)
Better bio-technologies (GRIN: Genomics, Robotics, Informatics, Nanotechnology)
No efficient technology for epigenomics, translocations or de novo mutations.
Algorithms for multi-locus association studies
Systems view of cancer by integrating data from multiple sources:
Genomic, Epigenomic, Transcriptomic, Metabolomic & Proteomic.
Regulatory, Metabolic and Signaling pathways

57. Mishra’s Mystical 3M’s Rapid and accurate solutions
Bioinformatic, statistical, systems, and computational approaches.
Approaches that are scalable, agnostic to technologies, and widely applicable
Promises, challenges and obstacles—
Measure
Mine
Model

58. Discussions
Q&A…

59. Answer to Cancer
“If I know the answer I'll tell you the answer, and if I don't, I'll just respond, cleverly.”
US Secretary of Defense, Mr. Donald Rumsfeld.

60. To be continued…
Break…