Structural Bioinformatics

1
Structural Bioinformatics

• Primary structure
• 20100 possibilities, but about 103-4 families
• High spatial locality
• Secondary structure
• Helix, Sheet, Loop, Coil
• Intermediate spatial locality
• Tertiary structure
• Only so many folds
• Evolutionary bias?
• Only so many structures? Smaller folding space?
• Low spatial locality
• Primary structure gt Secondary structure,
  Tertiary structure gt Function

2
Structural Bioinformatics

• Macromolecular complex gt
• Protein gt (No. of species 104)
• Fold gt (103)
• Domain gt (103-4)
• Module gt
• Motif gt
• Residue gt
• Atom

Level of abstraction
3
Primary structure

• Determined
• Experimentally
• Sequencing
• Computationally
• Proteome prediction from genome
• Finite number of real-world families based on
  sequence similarity
• Significance Sine qua none
• Databases Swiss-Prot, PIR, Genpept

4
Secondary structure

• Determined
• Experimentally
• Circular dichroism, NMR, Raman spectroscopy
• Computationally
• Sliding window context analysis
• Periodicity analysis
• Significance
• Higher order building block
• Mechanistic significance in protein folding

5
Tertiary structure

• Determined
• Experimentally
• X-ray crystallography, NMR
• Computationally
• Based on similarity to known structures (Homology
  modeling)
• a priori
• Significance
• Level of abstraction highly indicative of
  function
• Databases PDB, SCOP, CATH, FSSP

6
Structural representation

• Average (crystallography) or set (NMR) of
  structures specified
• Cartesian coordinates (x,y,z) for each atom in
  structure One vector for each atom
• Internal coordinate representation (edges,
  angles) Set of inter-atom vectors

7
Comparing structures

• Align structures
• Exhaustive search space vast (Complexity?)
• Algorithms typically use subset of coordinates
• Compute similarity
• Root mean square deviation
• Other normalized measures

8
Structural alignment

• Vector based strategies
• Compare summarized vector representations (VAST)
• Compare nearest-neighbor vectors (SSAP) by
  dynamic programming
• Distance matrix comparisons (a là Dot-matrix)
• Subset of internal coordinate space (DALI)

9
VAST alignment

• Use only subset of atom coordinates
• Replace atom coordinates with vector coordinates
  corresponding to secondary structural elements
• Compare sets of vectors to assess similarity

10
Double dynamic programming

• First level
• Represent each residue by neighborhood vector
• Compare n versus m neighborhood vectors
• Generate optimal alignment based on vector
  differences and dynamic programming
• Second Level
• Add matrix scores if paths cross in a cumulative
  matrix
• Generate optimal alignment based on the
  cumulative matrix

11
Distance matrix based alignment

• Generate dot-matrix of inter-residue distances,
  using threshold
• Pick out secondary structure elements based on
  matrix patterns
• Compare two matrices to generate structural
  alignment

12
Summary

• Sequence similarity (gt 50 identity) implies
  structural similarity. Converse not necessarily
  true (evolutionary convergence/information
  convergence)
• Structural similarity algorithms are heuristic
  ways to assess structural similarity
  independent of sequence similarity
• Structural variation is smaller that the number
  of possible sequences