Representing DNA Structure and Using Metaphors

1
Representing DNA Structure andUsing Metaphors
Analogies

• Srinivas Narayanan
• Vinay K. Chaudhri

2
Hypothesis

• Knowledge is grounded in small number of building
  blocks that are extended by composition and
  analogy

3
Our Work

• Investigate metaphors as an organizational
  principle in the knowledge base
• Orthogonal to taxonomic organization
• Represent about 20-30 metaphors grounded in
  processes and forces
• Use structural invariants as a basis for
  analogical inference propagation and validation

4
Contribution to E2E Goals

• A more natural organization for the KB
• Provide some of the building blocks for the core
  knowledge
• Be able to use structure mapping in the E2E system

5
Outline

• Representing DNA Structure
• Pump priming
• A first cut representation
• Metaphors used in DNA representation
• Chemical bond and connector
• Holding together
• Sample Questions
• Constructing RNA representation from DNA
• Using Analogy GUI from NWU

6
Outine
7
DNA Structure

• A DNA molecule consists of two long
  polynucleotide chains composed of four types of
  nucleotide subunits.
• The two chains are held together by hydrogen
  bonds between the base portions of the
  nucleotides. As we saw in Chapter 2, nucleotides
  are composed of a five-carbon sugar to which are
  attached one or more phosphate group.
• The nucleotides are covalently linked together in
  a chain through the sugars and phosphates, which
  thus form a backbone of alternating
  sugar-phosphate-sugar-phosphate.

8
Pump Priming

• Domain-specific concepts that should already be
  in the KB
• Nucleotide
• Chemical Bond
• Five Carbon Sugar
• Phosphate Group
• Covalent Bond
• Ribose and Deoxyribose
• Chemical Elements Nitrogen, Hydrogen, etc.

9
Representing Nucleotide
(every Nucleotide has (parts ((a
RingCompound with ((name ((Base)))
(parts ((a Nitrogen with (linkedTo
((the FiveCarbonSugar parts of
Self)))))))) (a
FiveCarbonSugar with (instance-of
((constraint ((set Ribose Deoxyribose)
includes
TheValue))))) (a PhosphateGroup with
(linkedTo ((the FiveCarbonSugar parts of
Self)))) (at-least 1 PhosphateGroup) (exactly
1 Base) (exactly 1 FiveCarbonSugar)))) (MELD
Translation is available)
10
Technical Issues

• SMEs must be able to specialize by adding
  additional constraints
• Simply connecting the components does not seem to
  be enough
• Possible Solutions
• GKB-style constraint editing
• Use dialogs to elicit constraints

11
Representing DNA Chains

• A DNA chain consists of Nucleotide element each
  of which has one of Adenine, Gyanine, Thymine,
  or Cytosine as a base
• Ends of the DNA chain have polarity (a 5 and a
  3 end).

12
Representing DNA Chain
(every DNA-Chain has ((prettyName (("DNA
Chain" "DNA Strand"))) (name ((DNA-Chain
DNA-Strand))) (chainEnd ((the1
chainElement of Self with ((name
((threeprimeEnd))) (parts ((a DeoxyRibose
with (exactly 0 connectedTo))))))
(the last chainElement of Self
with ((name ((fiveprimeEnd))) (parts ((a
PhosphateGroup with (exactly 0
connectedTo)))))))) (chainElement
((must-be-a a Nucleotide with
((constraint ((instance-of (set Adenine Gyanine
Cytosine Thymine))))) (orientation
((VectorFromToFn FiveMinuteEnd ThreeMinuteEnd)))))
)
13
Technical Issues

• Need a generic representation component modeling
  chains
• Need to state spatial direction

14
Representing DNA Molecule

• A DNA molecule has the shape of a double helix.
• A DNA molecule consists of a pair of DNA chains.
• The Chains are anti-parallel.
• The Chains are held together by chemical bonds
  between the complementary bases along each chain.
  The complementary bases are A-T and C-G. The
  bonds are AT-bond and CG bond.

15
Representing DNA Molecule
(every DNA-Molecule has ((shape
((DoubleHelix))) (parts ((seq ((a
DNA-Chain with ((chainElement
((a Nucleotide with ((parts ((a Base with
((location ((the interior of
Self))) (heldBy ((a ChemicalBond
with ((held ((Self)
(a chainElement of (the2 DNA-Chain part
of Self)))) constraint

((instance-of

(set AT-Bond GC-Bond))))))))
(exactly 1 heldby)))
..
16
Technical Issues

• Need representation components for Hold Together.
• Need representation of Bonds.

17
Representing Holds

• Holding objects in place involves the application
  of a holding force which maintains the relative
  position and orientation of the objects.
• Holding Together is a hold whose holding force is
  directed inward (relative to the objects). The
  tendency of the objects is to move apart.
• Holding Apart is a hold whose holding force is
  directed outward (relative to the objects). The
  tendency of the objects is to move toward each
  other.

18
Representing Hold
(every Hold has
((holdingForce ((a Force with
((experiencer ((the patient of Self)))))))
(patient ((a PartiallyTangible))) (maint
ainsInPosition ((the experiencer of
Self))) (pcs-list ((triple (forall (the held
of Self)) location ?loc) (triple (forall (the
held of Self)) orientation ?orientation)
(triple (the magnitude of holdingForce) gt
0)))))
Issue How is maintainsInPosition represented?
19
Variations of Hold
(every HoldTogether has ((patient ((seq
(a PartiallyTangible with
(potentialDirectionOfMotion
((VectorFromToFn (the2 patient of Self) (the1
patient of Self))))) (a
PartiallyTangible with
(potentialDirectionOfMotion
((VectorFromToFn (the1 patient of Self) (the2
patient of Self))))))) (holdingForce ((a
Force with (composedOf
((a Force with ((direction ((VectorFromToFn

(the1 patient of Self) (the2 patient of
Self)))) (experiencer
(the1 patient of Self)))) (a Force with
((direction ((VectorFromToFn
(the2 patient of
Self) (the1 patient Of Self)))) .
OTHER VARIATIONS HOLD APART, HOLD AT A
DISTANCE, HOLD IN
20
Technical Issues
Need a general structure that captures more
complicated force interactions and processes.
21
Chemical Bonds as Connectors
Instance of the General Metaphor Mapping Forces
gt Force-bearing Objects Chemical Bonds are
broken, created, fragile, weak, destroyed, etc.
(every Connector has (primaryFunction
((HoldTogether))) (ChemicalBond has
((construedAs ((Connector)))))
22
Representing Metaphors in the Knowledge Base

• (every DomainMap has
• ((source ((a Thing)))
• (target ((a Thing)))))
• (ConnectorBondMap with
• (source ((a Connector)))
• (target ((a ChemicalBond with
• ((primaryFunction ((the primaryFunction
  of Self))))

23
Sample Questions

• The nucleotide sequence of one DNA strand of a
  DNA double helix is 5-GGATTTTCCCAAGG-3. What
  is the sequence of the other strand?
• Follows from the structure of DNA and the
  definition of AT and CG bonds
• Need to figure out if KM inference engine will do
  this inference

24
Sample Questions

• Which of the following DNAs would melt first?
• 5-GCGGGGCCAGCCT-3
• 3-CGCCCCGGTCGGA-5
• 5-AAATTTTAAAAGAAA-3
• 3-TTTAAAATTTTCTTT-5
• Involves use of problem solving knowledge
• Need to represent qualitative relations

25
Outline

• Representing DNA Structure
• Pump priming
• A first cut representation
• Metaphors used in DNA representation
• Chemical bond and connector
• Holding together
• Sample Questions
• Constructing RNA representation from DNA
• Using Analogy GUI from NWU

26
INITIAL SCREEN
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
Original Individuals
New Individuals
Original Statements
New Statements
Correspondences
Candidate Inferences
New
Delete
Accept
27
BASE LOADED
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
New Individuals
Original Individuals

1
DNA
2
AT- Bond
3
CG-Bond
4
DNA-Chain
5
Phosphodigester Linkage
Original Statements
New Statements

7
DNA HAS SHAPE Double Helix
DNA HAS PART DNA-Chain
9
19
DNA-Chain HAS PART Nucleotide
21
Nucleotide HAS PART Deoxyribpse
Correspondences
Candidate Inferences
New
Delete
Accept
28
TARGET WITH INDIVIDUALS
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
Cell
RNA
New Individuals
Original Individuals


1
DNA
66
RNA
2
AT- Bond
67
AU- Bond
3
CG-Bond
69
CG- Bond
4
DNA-Chain
70
RNA-Chain
5
Phosphodigester Linkage
74
Phosphodigester Linkage
Original Statements
New Statements


7
DNA HAS SHAPE Double Helix
DNA HAS PART DNA-Chain
9
19
DNA-Chain HAS PART Nucleotide
21
Nucleotide HAS PART Deoxyribpse
Correspondences
Candidate Inferences
N Original Individual
N New Individual
New
Delete
Accept
29
ADDING TARGET ISA STATEMENTS
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
Cell
RNA
New Individuals
Original Individuals


1
DNA
66
RNA
2
AT- Bond
67
AU- Bond
3
CG-Bond
69
CG- Bond
4
DNA-Chain
70
RNA-Chain
5
Phosphodigester Linkage
74
Phosphodigester Linkage
Original Statements
New Statements


7
DNA HAS SHAPE Double Helix
33
RNA ISA Chemical Compound
DNA HAS PART DNA-Chain
9
RNA-Chain ISA Chain
34
19
DNA-Chain HAS PART Nucleotide
21
Nucleotide HAS PART Deoxyribpse
Correspondences
Candidate Inferences
N Original Individual
N New Individual
New
Delete
Accept
30
SUGGESTING CORRESPONDENCES
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
Cell
RNA
New Individuals
Original Individuals


1
DNA
66
RNA
2
67
AU- Bond
AT- Bond
3
69
CG- Bond
CG-Bond
4
DNA-Chain
70
RNA-Chain
5
Phosphodigester Linkage
74
Phosphodigester Linkage
Original Statements
New Statements


7
DNA HAS SHAPE Double Helix
33
RNA ISA Chemical Compound
DNA HAS PART DNA-Chain
9
RNA-Chain ISA Chain
34
19
DNA-Chain HAS PART Nucleotide
21
Nucleotide HAS PART Deoxyribpse
Correspondences
Candidate Inferences
N Original Individual
N New Individual
New
1 DNA
66 RNA
Delete
Accept
31
RESULTS OF INITIAL MATCHING
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
Cell
RNA
New Individuals
Original Individuals


1
DNA
66
RNA
2
67
AU- Bond
AT- Bond
3
69
CG- Bond
CG-Bond
4
DNA-Chain
70
RNA-Chain
5
Phosphodigester Linkage
74
Phosphodigester Linkage
Original Statements
New Statements


7
DNA HAS SHAPE Double Helix
78
RNA HAS PART RNA-Chain
DNA HAS PART DNA-Chain
9
82
RNA -Chain HAS PART Nucleotide
19
DNA-Chain HAS PART Nucleotide
Nucleotide HAS PART Deoxyribose
57
21
Nucleotide HAS PART Deoxyribpse
RNA HAS SHAPE Double Helix
48
Correspondences
Candidate Inferences
N Original Individual
N New Individual
New
1 DNA
66 RNA
Delete
2 AT-Bond
67 AU-Bond
3 CG-Bond
69 CG-Bond
Accept
4 DNA-Chain
70 RNA-Chain
32
DELETING CORRESPONDENCES
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
Cell
RNA
New Individuals
Original Individuals


1
DNA
66
RNA
2
67
AU- Bond
AT- Bond
3
69
CG- Bond
CG-Bond
4
DNA-Chain
70
RNA-Chain
5
Phosphodigester Linkage
74
Phosphodigester Linkage
Original Statements
New Statements


7
DNA HAS SHAPE Double Helix
78
RNA HAS PART RNA-Chain
DNA HAS PART DNA-Chain
9
82
RNA-Chain HAS PART Nucleotide
19
DNA-Chain HAS PART D-Nucleotide
Nucleotide HAS PART Deoxyribose
57
21
Nucleotide HAS PART Deoxyribpse
RNA HAS SHAPE Double Helix
48
Correspondences
Candidate Inferences
N Original Individual
N New Individual
New
1 DNA
66 RNA
Delete
2 AT-Bond
67 AU-Bond
3 CG-Bond
69 CG-Bond
Accept
4 DNA-Chain
70 RNA-Chain
33
ADDING TARGET KNOWLEDGE
OLD CASE
NEW CASE
Target Case
Domain
Base Case
Domain
DNA
Cell
Cell
RNA
New Individuals
Original Individuals


1
DNA
66
RNA
2
67
AU- Bond
AT- Bond
3
69
CG- Bond
CG-Bond
4
DNA-Chain
70
RNA-Chain
5
Phosphodigester Linkage
74
Phosphodigester Linkage
Original Statements
New Statements


7
DNA HAS SHAPE Double Helix
78
RNA HAS PART RNA-Chain
DNA HAS PART DNA-Chain
9
82
RNA-Chain HAS PART Nucleotide
19
DNA-Chain HAS PART D-Nucleotide
RNA-Chain HAS SHAPE Chain
99
21
Nucleotide HAS PART Deoxyribpse
101
R-Nucleotide HAS PART Ribose
Correspondences
Candidate Inferences
N Original Individual
N New Individual
New
1 DNA
66 RNA
Delete
3 CG-Bond
69 CG-Bond
4 DNA-Chain
70 RNA-Chain
Accept
34
Conclusion

Representing the structure of DNA draws building
blocks from space, force dynamics and event
structure metaphors. RNA structure can be
specified using a within-domain analogy with
DNA. We have a design and are working on
an implementation of the relevant concepts.