Methods of Modeling Protein Structure

1
Methods of ModelingProtein Structure

• Peter Bell
• December 6, 2002

2
3-D Structure of Proteins

• Structure of natural proteins guided by 2 sets of
  principles on vastly different time scales
• Laws of Physics
• Chemical bonds, H-bonds, electonic interactions
• Theory of Evolution
• Families of similar proteins share similar
  sequences, structure, and often function

3
Modeling Proteins

• Two general methods to modeling
• Each based on either of the principles governing
  3-D structure
• Comparative Modeling based on similarities from
  evolution
• De novo or ab initio based on physicsenergies

4
Comparative Methods

• Four Main Steps in this process
• Finding known structures related to the target
  sequence being modeled. (finding templates)
• Aligning sequence with known structures
• Building a model
• Assesing the model

5
Comparative Modeling

• Finding Templates
• 3-D structures are more conserved that primary
  sequences through evolution
• Proteins with similar sequences will often share
  structural characteristicsfind using sequence
  comparison methods such as PSI-BLAST
• Sometimes, very different sequences will share
  structurefind these using the Protein Data Bank
  (makes use of coarse structures for quick
  comparisons)

6
Comparative Modeling

• Building the model
• After finding similar sequences, a structure is
  modeled using that related sequence.
• Core regions are kept intact (helices, sheets)
• Loops and side chains using various methods to
  minimize energies in the presence of the core
  regions.
• Three main methods
• Loops and side chains based on alignment to
  related sequences in other proteins
• Approx. position of atoms is calculated from
  approx. position of conserved atoms in template.
• Distance Geometry or spatial restraints are
  optimized

7
Comparative Modeling

• Assesing the Model
• Accuracy associated with the percent of the
  sequence that matches the template
• Most errors in side chains or in loops
• These are mostly insignificant as these parts
  dont effect function.
• Other errors include small shifts and distortions
  in other parts of the protein
• Error increases rapidly below 30 sequence
  similarity

8
De novo Prediction

• Can be done with any proteinno other members of
  the family need be known
• Based on assumption Native State is at the
  global free energy minimum
• A large scale search is carried out to find this
  minimum with tertiary structure.

9
De novo Prediction

• Two key components in finding minimum
• Methods for carrying out search efficiently
• Free energy function used to calculate energy
• Complexity of these increases as computing power
  increases
• Often only a subset of actual atoms in a protein
  is modeled (leave out hydrogen, for example)
• The potential functions must then account for
  these absent atoms.
• Some newer methods flip certain sequences into
  possible secondary structures, calculate energy,
  then find how these fit into tertiary structure

10
De novo Prediction

• Accuracy is much lower than comparative methods
  (with gt30 sequence similarity)
• Basic structure can still be determined
  reasonably well with 150 amino acids or less
• Accuracy too low if high-resolution info is
  needed.
• Low-resolution info is useful in finding
  structural and functional relationships no
  apparent by comparison of AA sequence.

11
Summary

• Two methods available to model proteins based on
  their AA sequence
• Comparative Theory of Evolution
• De novo Laws of Physics
• Comparative more accuratebut requires other
  structures to be known
• De novo can be used for any protein but requires
  more computing power and wont produce highly
  accurate models.

12
References

• Baker, D. Sali, A. Protein Structure Prediction
  and Structural Genomics. Science 2001, 294,
  93-96.
• Fiser, A. Feig, M. Brooks, C. L. Sali, A.
  Evolution and Physics in Comparative Protein
  Structure Modeling. Acc. Chem. Res. 2002, 35,
  413-421.