Protein Structure Prediction
CHOU FASHMAN METHOD
• Secondary protein structure prediction
• Based on propensity value
• Tendency of the amino acids to behave more in alpha helix and beta sheets
• Propensity value for alpha helix =
(Frequency of selected amino acid to be in alpha helix / Frequency of
residues to be in alpha helix)
Rules for secondary structure prediction
1. Helix nucleation
Randomly find 6 stretch of sequence
> 1/3 helix breaker – No alpha helix
< 1/2 helix maker – No alpha helix
2. Helix termination
Propensity value (Pα) < 100 for continuous 4 amino acid – No alpha helix form
N terminal – charged amino acid has tendency to present – (-ve)
C terminal – (+ve) charged amino acid
For Beta sheet
1. Helix nucleation
Randomly find 5 stretch of sequence
> 1/3 beta sheet breaker – No beta sheet
< 1/2 beta sheet maker – No beta sheet
2. Helix termination
Propensity value (Pβ) < 100 for continuous 4 amino acid – No beta sheet form
N terminal – charged amino acid has tendency to present – (+ve)
C terminal – (-ve) charged amino acid
Tertiary Protein Prediction
Homology Modeling
• Computational method – comparative model of protein prediction
structure
• Construct unknown protein structure by using the template protein
• Template – protein database
• Compare target protein with template protein by sequence similarity
• Sequence similarity greater than or equal to 30% - Perform homology
modeling
Workflow
• Identify related structures
(Templates)
• Select template
• Align target sequence with
template structures
• Build a model for the target
(using information from the
template structure)
• Evaluate the model
• Model ok
Steps
• Template recognition and Initial
alignment
• Alignment correction
• Backbone generation
• Loop modeling
• Side chain modeling
• Model optimization
• Model validation
Template recognition and Initial alignment
• Search homologous proteins for determined structures
• Sequence alignment – BLAST & FASTA
• Generated Hits – scoring matrix – Residual exchange matrix and
alignment matrix
Alignment correction
• Difficult to align two sequences where sequence identity is very low
• Use other sequences from homologous protein to find a solution
Backbone generation
• When alignment correction is done perfectly, backbone of the target
can be found
• Coordinates of the template sequence is copied to the target
sequence
• If some residues are identical, the side chain coordinates are also
copied
• For eg. If amino acid proline present in 12th position of template
sequence is also found in the same position of target sequence then
the residue can be copied directly to the target sequence
Loop modeling
• After sequence alignment, there are often regions created by
insertions and deletions that leads to gaps in alignment
• These gaps are modelled by loop modeling
• Search loops from the database and superimpose over the target
structure - the database searching method
• Generates many random loops and searches for one that has very low
energy, φ and ψ angles in the allowable regions of Ramachandran
plot - the Ab – initio method
Side chain modeling
• Estimate protein-ligand interaction at active sites
• Protein-protein interaction at the contact interface
• Side chain will be generated with the help of confirmational changes
and torsion angle of side chains
• Best torsion angle with the lowest energy with neighboring atom will
be taken into consideration
• Rotamer library is used to get the side chain of the target protein
sequence by comparing known protein sequence
Model optimization
• Energy minimization will be done
• Main goal is to release steric collision without altering the structure
• Model optimization done – Molecular dynamic simulation done
where the atoms tend to move to the global minima by considering
conditions like heating and cooling which get true protein structure
• Energy bonding
• Stretching energy
• Bending energy
• Torsion energy
• Non-bonded interaction energy
Model validation
• Every model tends to have errors
• Two main reasons
• The percentage sequence identity between template and target. If its > 90% then the model
can be compared with the crystallographic determined structures. If it is < 30% then error
occurs
• The no. of errors in template
• Final model evaluated for φ, ψ angles, chirality, bond lengths, close contact, and
other stereochemical properties
• Modeling programs
• Modeller
• SWISS MODEL
• Schrodinger
• 3D-JIGSAW
• Successful model depends on
• Template selection
• Algorithm used
• Model validation
Zones of sequence alignments
Example
Advantages
• Find the location of the alpha chain in the folded protein
• Study mutagenesis experiments and structure-function relationships
• Find active sites, loops and side chains in the conserved region of the
protein
Disadvantages
• Unable to predict confirmations of insertions and deletions or side
chain positions at high level of accuracy
• Not useful for modeling ligand docking studies for drug designing and
development process
• If template and target identity is > 70% it can be used for the
designing purposes
Fold recognition method / Threading
