Structural Bioinformatics Lecture 1: Protein Structure Prediction by
Comparative Modelling
Homology Modeling-
Outline:
Definition,
Steps,
Diagram,
Uses
Protein structure prediction is an important aspect of computational biology that aims to determine
the three-dimensional (3D) structure of a protein. Homology modeling is one of the computational
methods of protein structure prediction.
Homology modeling, also known as comparative modeling, is a method used to predict the
3D structure of a protein with an unknown structure by using the known structure of a
homologous protein.
Homology Modeling
The method relies on the fact that the 3D structure of proteins is often better conserved than their
amino acid sequence. Therefore, proteins with similar sequences are likely to have similar
structures.
This method is becoming increasingly important as the number of experimentally determined
structures is limited compared to the number of protein sequences available. As the gap between
the number of protein sequences and experimentally determined structures continues to grow,
homology modeling has an increasingly important role to play in filling this gap.
Steps in Homology Modeling
Steps in Homology Modeling
The overall homology modeling procedure consists of a series of steps:
1. Identification and Selection of Template Structures:
The first step in this process is identifying and selecting appropriate template structures from the
Protein Data Bank (PDB).
To identify appropriate templates, pair-wise sequence alignment methods such as BLAST can
be used to identify proteins with high sequence homology to the query protein in the PDB.
More sensitive methods involve performing multiple sequence alignments or using automated
iterative methods such as PSI-BLAST.
If the homology between the protein of interest and the potential templates is low (below 35%),
alternative methods such as profile-profile alignments, and Hidden Markov Models (HMMs) are
used to reduce gaps in the alignment.
Once several templates are identified, the most appropriate ones for modeling must be chosen.
Factors such as sequence identity and similarity, alignment scores, and phylogenetic
relationships should be considered, as well as other information such as biological function and
environmental context.
2. Alignment of the Query Sequence with the Template:
In homology modeling, the alignment of the query sequence with the template is an important
step in the process.
It is essential to achieve a high-quality sequence alignment as it directly impacts the accuracy of
the resulting model. A homology model can only be as good as the sequence alignment.
Multiple sequence alignment is preferred as it contains evolutionary information that is
important for structure and function. Effective multiple sequence alignment programs like
Proline and T-Coffee are recommended for use.
However, even the best program may contain errors, so it is necessary to visually examine the
alignment to confirm that conserved residues are aligned correctly.
In case of inaccuracies, manual adjustments may be necessary, including repositioning gaps,
changing substitution scores, or moving sequence segments to improve alignment quality.
3. Building a Three-Dimensional Model of the Query Protein:
Sequence alignment is followed by the 3D model building of the target proteins.
There are different methods used for generating 3D models of protein based on its templates.
These methods are classified into four categories: rigid-body assembly, segment matching,
spatial restraint, and artificial evolution methods.
In rigid-body assembly, the protein structure is broken down into basic conserved core regions,
loops, and side chains. The rigid body parts are picked up from the template protein structures
and brought together using tools like 3D-JIGSAW, BUILDER, and SWISS-MODEL.
Segment matching involves using a cluster of atomic positions from template structures as
leading positions. These atomic positions are used to select segments from known structures in
the database based on sequence identity, geometry, and energy. These segments are aligned with
the template structure to create a model of the target protein. This can be done with
SEGMOD/ENCAD.
The spatial restraint method builds protein structures by applying spatial restraints to guide
the building of a model that closely matches the template structure. The restraints are determined
by stereochemical principles on bond length, bond angle, dihedral angles, and van der Waals
contact distances. MODELLER is a software package commonly used to perform this method.
The artificial evolution method involves using the rigid-body assembly method and stepwise
template evolutionary mutations together until the template sequence is the same as the target
sequence. This can be performed with NEST.
4. Loop Modeling
There are often gaps or insertions that can occur in sequence alignments when modeling proteins.
These gaps are called loops and are structurally less conserved during evolution.
Loop modeling is an important step in protein structure prediction because loops play a crucial
role in determining the function of a protein. However, predicting their structure is a complex
process.
There are two main methods for loop prediction: the database search method and the
conformational search method.
The database search method involves comparing the target protein sequence with all known
protein structures to identify segments that can be used to model the loop.
The conformational search approach involves optimizing a scoring function to find the best
possible loop conformation. This method generates many random loops and searches for the one
that fits the target protein structure. This approach is also called ab initio because it does not rely
on pre-existing structural information.
There are specialized programs available for loop modeling, such as FREAD, PETRA, and
CODA.
5. Side Chain Modeling
Side chain modeling is the process of predicting the conformation of the side chains of amino
acids in a protein structure. This process is crucial in evaluating protein-ligand and protein-
protein interactions.
This is usually done by placing the side chains onto the backbone coordinates that are derived
from a parent structure or from ab initio modeling simulations.
Searching every possible conformation of a side chain is computationally time-consuming and
not effective so, most side chain prediction programs use the preferred conformation called
rotamers. These rotamers are stored in a rotamer library, which is a collection of preferred side
chain conformations ranked by their frequency of occurrence.
Different energy functions and search strategies are used to select the most appropriate rotamer
for each amino acid side chain based on the preferred protein sequence and the given backbone
coordinates.
There are various tools available for side chain modeling, such as RAMP and SCWRL.
6. Model optimization
The next step is to optimize the model. Optimization of a model refers to the process of refining
the initial model to improve its accuracy and reliability.
This optimization process involves adjusting the positions of the atoms in the model to reduce
any clashes or steric hindrances between atoms.
Energy minimization process is one of the methods of model optimization. It calculates the
potential energy of the model and adjusts the positions of the atoms to minimize this energy.
Excessive energy minimization can often move residues away from their correct positions, so
only limited energy minimization is recommended.
In addition to energy minimization, other optimization techniques such as molecular dynamics
and Monte Carlo simulations can be used.
Molecular dynamics simulations involve simulating the movements and interactions of the
atoms in the model over time, which allows the exploration of different conformations and the
identification of the most stable or energetically favorable ones.
Monte Carlo simulations use random sampling to explore different configurations of the
molecule and identify the most stable or energetically favorable ones.
7. Model Validation and Evaluation
After the 3D model of the query protein has been built, it is essential to validate and evaluate its
quality to ensure that it is biologically relevant and can be used for further studies.
This involves comparing the model to real experimental 3D structures and using various
evaluation methods to assess the model’s stereochemistry, physical parameters, knowledge-
based parameters, statistical mechanics, and other criteria.
Some commonly used programs for evaluating protein structures are WHATIF, PROCHECK,
and PROSA. These programs compare the properties of the homology model to those found in
the experimental structures in the PDB.
The quality of the homology model largely depends on the accuracy of the sequence alignment
and the parts of the protein where the template is not similar to the query protein.
Applications of Homology Modeling
Homology modeling has many practical applications:
Homology modeling is important for the prediction of protein structures that are difficult to
obtain experimentally using techniques like X-ray crystallography or NMR.
Homology modeling is used to predict the functional properties of proteins with unknown
structures by modeling the protein structure and comparing it with known structures of related
proteins.
Homology modeling is also useful in studying the evolution of protein families by comparing
the structures of related proteins from different species to identify the changes that have occurred
during evolution.
Homology modeling can be used to identify potential drug targets by predicting the structure
of a protein to understand its function and identify potential sites for drug binding. This
information can be used to design new drugs or optimize existing ones.