The document discusses protein structure modeling through homology modeling. It describes the key steps in homology modeling which include: (1) finding a suitable template through database searches, (2) aligning the target sequence to the template, (3) assigning coordinates from conserved regions of the template, (4) building loops and variable regions either from other structures or de novo, (5) searching for optimal side chain conformations, and (6) refining the model through molecular mechanics. The document emphasizes validating the final model to identify any inherent errors from the template or modeling process.

1. Protein Structure Modeling

2. Learning Objectives
• At the end of this topic the student should be
able to:
(i) Perform computer aided- protein structure
determination and validation
(ii) Design docking experiments
(iii) Demonstrate understanding of protein
interactions and simulations

3. Computer-Aided Protein Structure
Prediction
Protein
Sequence +
Structure

4. Recap
Structure prediction aims at building models of 3D structures of proteins that
could be useful for understanding structure function relationships.

5. STAT115 5
Protein Structure Levels
Folded structure:
lowest energy state

6. Watch Video
http://www.youtube.com/watch?v=fvBO3TqJ6F
E&feature=fvw

7. Protein Structure Determination &
Prediction
• Experimental Techniques
– X-ray Crystallography
– NMR
• Limitations of Current Experimental
Techniques
– Protein DataBank (PDB) -> 30,000 protein structures
– Unique structure are between 4000 to 5000 only
– Non-redudant (NR) -> 10,000,000 proteins sequences
• Importance of Structure Prediction
– Fill gap between known sequence and structures
– Protein Engineering.- To alter function of a protein
– Rational Drug Design

8. STAT115 8
Protein Structure Prediction
• Since AA sequence determines structure, can we
predict protein structure from its AA sequence?
– predicting the three angles, unlimited DoF!
• Physical properties that determine fold
– Rigidity of the protein backbone
– Interactions among amino acids, including
• Electrostatic interactions
• van der Waals forces
• Volume constraints
• Hydrogen, disulfide bonds
– Interactions of amino acids with water

9. Why Protein Modeling
 The goal of research in the area of structural genomics is to
provide the means to characterize and identify the large
number of protein sequences that are being discovered
Knowledge of the three-dimensional structure
 helps in the rational design of site-directed mutations
 can be of great importance for the design of drugs
 greatly enhances our understanding of how proteins function
and how they interact with each other , for example, explain
antigenic behaviour, DNA binding specificity, etc

10. Computational methods for Protein Structure
Prediction
 Homology or Comparative Modeling
 Fold Recognition or threading Methods
 Ab initio methods that utilize knowledge-based information
 Ab initio methods without the aid of knowledge-based
information

11. Why do we need computational approaches?
 Structural information from x-ray crystallography or NMR
results
obtained much more slowly.
techniques involve elaborate technical procedures
many proteins fail to crystallize at all and/or cannot be
obtained or dissolved in large enough quantities for NMR
measurements
The size of the protein is also a limiting factor for NMR
 With a better computational method this can be done extremely fast.

12. Tertiary Structure Prediction Methods
Any given protein sequence
Structure selection
Compare sequence identity with proteins that have solved structure
Homology
Modeling
> 35%
Fold
Recognition
ab initio
Folding
< 35%
< 35%
Structure refinement
Final Structure
Structure selection

13. Homology Modeling
• Predicts the three-dimensional structure of a
given protein sequence (TARGET) based on an
alignment to one or more known protein
structures (TEMPLATES)
• If similarity between the TARGET sequence and
the TEMPLATE sequence is detected, structural
similarity can be assumed.
• In general, 30% sequence identity is required for
generating useful models.

14. Homology Modeling Process
 Template recognition
 Alignment
 Determining structurally conserved regions
 Backbone generation
 Building loops or variable regions
 Conformational search for side chains
 Refinement of structure
 Validating structures

15. Template Recognition
• First we search the related proteins sequence(templates) to the target
sequence in any structural database of proteins
• The accuracy of model depends on the selection of a proper template
• FASTA and BLAST (PSI-BLAST) from EMBL-EBI and NCBI can be used
• This gives a probable set of templates but the final one is not yet decided
• After intial aligments and finding structurally conserved regions among
templates, we choose the final template

16. How good can homology modeling be?
Sequence Identity
60-100% Comparable to medium resolution NMR
Substrate Specificity
30-60% Molecular replacement in crystallography
Support site-directed mutagenesis
through visualization
<30% Serious errors!!!

17. Protein Structural Databases
• Templates can be found using the TARGET sequence as a
query for searching using FASTA or BLAST
– PDB (http://www.rcsb.org/pdb)
– MODELLER
(http://guitar.rockefeller.edu/modeller/modeller.html)
– ModBase (http://pipe.rockefeller.edu/modbase/general-
info.html)
– 3DCrunch
(http://www.expasy.ch/swissmod/SM_3DCrunch.html)

18. Target-Template Alignment
• No current comparative modeling method can
recover from an incorrect alignment
• Use multiple sequence alignments as initial guide.
• Consider slightly alternative alignments in areas of
uncertainty, build multiple models
Note: sequence from database versus sequence from
the actual PDB are not always identical

19. Target-Multiple Template Alignment
• Alignment is prepared by superimposing all template
structures
• Add target sequence to this alignment
• Compare with multiple sequence alignment and adjust

20. Gaining confidence in template searching
• Once a suitable template is found, it is a good idea to do a
literature search (PubMed) on the relevant fold to
determine what biological role(s) it plays.
• Does this match the biological/biochemical function that
you expect?

21. Other factors to consider in selecting
templates
• Template environment
– pH
– Ligands present?
• Resolution of the templates
• Family of proteins
– Phylogenetic tree construction can help find the
subfamily closest to the target sequence
• Multiple templates?

22. Determining Structurally Conserved Regions
(SCRs)
• When two or more reference protein structures are available
• Establish structural guidelines for the family of proteins under consideration
• First step in building a model protein by homology is determining what
regions are structurally conserved or constant among all the reference
proteins
• Target protein is supposed to assume the same conformation in conserved
regions

23. Structurally Conserved Regions
 SCRs are region in all proteins of a particular family that
are nearly identical in structures.
 Tend to be at inner cores of the proteins
 Usually contains alpha-helices and beta sheets
 No SCR can span more than one secondary structure

24. Alignment of Target Protein with SCR
 After doing alignments and finding SCRs
 We align the unknown sequence with the aligned reference proteins with the
knowledge of SCRs
 Since SCR cant contain insertions and deletions
 The enhancement of standard alignment algorithm is used
 After finding the suitable template and aligning the unknown sequence with the
template
 Assignment of coordinates within conserved regions is done

25. Assignment of coordinates within conserved
region
 Once the correspondence between amino acids in the reference and model
sequences has been made, the coordinates for an SCR can be assigned
 The reference proteins' coordinates are used as a basis for this assignment
 Where the side chains of the reference and model proteins are the same at
corresponding locations along the sequence, all the coordinates for the amino
acid are transferred
 Where they differ, the backbone coordinates are transferred , but the side
chain atoms are automatically replaced to preserve the model protein's
residue types

26. Assignment of coordinates in loop or variable
region
Two main methods
 Finding similar peptide segments in other proteins
 Generating a segment de-novo

27. Finding similar peptide segment in other proteins
 Advantage: all loops found are guaranteed to have reasonable
internal geometries and conformations
 Disadvantage: may not fit properly into the given model protein’s
framework
In this case, de-novo method is advisable
Assignment of coordinates in loop or
variable region

28. Selection Of Loops
 Check the loops on the basis of steric overlaps
 A specified degree of overlap can be tolerated
 Check the atoms within the loop against each other
 Then check loop atoms against rest of the protein’s atoms

29. Sidechains on surface of protein
• Exposed sidechains on surface can be highly flexible without a
single dominant conformation
• So ultimately if these solvent exposed sidechains do not form
binding interactions with other molecules or involved in say, a
catalytic reaction, then accuracy may not be

30. Side Chain Conformation Search
 With bond lengths, bond angles and two rotable backbone bonds per
residue φ and ψ, its very difficult to find the best conformation of a side
chain
 In addition, side chains of many residues have one or more degree of
freedom.
 Hence Side chain conformational search in loop regions is must
 Side chain residues replaced during coordinate transformations should
also be checked

31. Selection Of Good Rotamer
 Fortunately, statstical studies show side chain adopt only a small number
of many possible conformations
 The correct rotamer of a particular residue is mainly determined by local
environment
 Side chain generally adopt conformations where they are closely packed

32. Refinement of model using Molecular
Mechanics
Many structural artifacts can be introduced while the model protein is
being built
 Substitution of large side chains for small ones
 Strained peptide bonds between segments taken from difference
reference proteins
 Non optimum conformation of loops

33. Model Validation
 Every homology model contains errors. Two main reasons
 % sequence identity between reference and model
 The number of errors in templates
 Hence it is essential to check the correctness of overall fold/ structure,
errors of localized regions and stereochemical parameters: bond
lengths, angles, geometries

34. Structural Prediction by Homology Modeling
Structural Databases
Reference Proteins
Conserved Regions Protein Sequence
Predicted Conserved Regions
Initial Model
Structure Analysis
Refined Model
SeqFold,Profiles-3D, PSI-BLAST, BLAST & FASTA, Fold-recognition methods
Cα Matrix Matching
Sequence Alignment
Coordinate Assignment
Loop Searching/generation
WHAT IF, PROCHECK, PROSAII,..
Sidechain Rotamers
and/or MM/MD
MODELER

35. Errors in Homology Modeling
a) Side chain packing b) Distortions and shifts c) no template

36. Errors in Homology Modeling
d) Misalignments e) incorrect template
Marti-Renom et al., Ann. Rev. Biophys. Biomol. Struct., 2000, 29:291-325.

37. Automated Web-Based Homology Modelling
 SWISS Model : http://www.expasy.org/swissmod/SWISS-MODEL.html
 WHAT IF : http://www.cmbi.kun.nl/swift/servers/
 The CPHModels Server : http://www.cbs.dtu.dk/services/CPHmodels/
 3D Jigsaw : http://www.bmm.icnet.uk/~3djigsaw/
 SDSC1 : http://cl.sdsc.edu/hm.html
 EsyPred3D : http://www.fundp.ac.be/urbm/bioinfo/esypred/
 HHpred server: https://toolkit.tuebingen.mpg.de/tools/hhpred

38. Tools for Model Evaluation
 WHAT IF http://www.cmbi.kun.nl/gv/servers/WIWWWI/
 SOV http://predictioncenter.llnl.gov/local/sov/sov.html
 PROVE http://www.ucmb.ulb.ac.be/UCMB/PROVE/
 ANOLEA http://www.fundp.ac.be/pub/ANOLEA.html
 ERRAT http://www.doe-mbi.ucla.edu/Services/ERRATv2/
 VERIFY3D http://shannon.mbi.ucla.edu/DOE/Services/Verify_3D/
 BIOTECH http://biotech.embl-ebi.ac.uk:8400/
 ProsaII http://www.came.sbg.ac.at
 WHATCHECK http://www.sander.embl-heidelberg.de/whatcheck/

39. Challenges
 To model proteins with lower similarities( eg < 30% sequence identity)
 To increase accuracy of models and to make it fully automated
 Improvements may include simulataneous optimization techniques in
side chain modeling and loop modeling
 Developing better optimizers and potential function, which can lead
the model structure away from template towards the correct
structure
 Although comparative modelling needs significant improvement,
it is already a mature technique that can be used to address
many practical problems

40. Comparative Modeling Server & Program
 COMPOSER
http://www.tripos.com/sciTech/inSilicoDisc/bioInformatics/matchmaker.h
tml
 MODELER http://salilab.org/modeler
 InsightII http://www.msi.com/
 SYBYL http://www.tripos.com/

41. STAT115 41
Fold Recognition
• Also called protein threading
• Given new sequence and library of known
folds, find best alignment of sequence to each
fold, returned the most favorable one

42. STAT115 42
Protein Folding with
Dynamic Programming
• D. Eisenburg 1994
• Align sequence to each fold (a string of
environmental classes)
• Advantages: fast and works pretty well
• Disadvantages: do not consider AA contacts

43. STAT115 43
Fold Recognition
• Each predictor can submit N top hits
• Every predictor does well on something
• Common folds (more examples) are easier to
recognize

44. STAT115 44
Fold Recognition
• Alignment (seq to fold) is a big problem

45. STAT115 45
ab initio
• Predict interresidue contacts and then
compute structure (mild success)
• Simplified energy term + reduced search space
(phi/psi or lattice) (moderate success)
• Creative ways to memorize sequence 
structure correlations in short segments from
the PDB, and use these to model new
structures: ROSETTA

46. 46
Ab initio prediction of protein structure
sample conformational space such that native-like conformations are found
astronomically large number of
conformations
5 states/100 residues = 5100 = 1070
select
hard to design functions
that are not fooled by
non-native conformations
(“decoys”)

47. Programs for Model Protein Prediction
https://molbiol-tools.ca/Protein_tertiary_structure.htm