Predicting Protein Structure: Comparative Modeling (homology modeling)

? KQFTKCELSQNLYDIDGYGRIALPELICTMF HTSGYDTQAIVENDESTEYGLFQISNALWCK SSQSPQSRNICDITCDKFLDDDITDDIMCAK KILDIKGIDYWIAHKALCTEKLEQWLCEKE Predicting Protein Structure: Comparative Modeling (formerly, homology modeling) Use as template & model 8lyz 1alc KVFGRCELAAAMKRHGLDNYRGYSLGNWVCAAK FESNFNTQATNRNTDGSTDYGILQINSRWWCND GRTPGSRNLCNIPCSALLSSDITASVNCAKKIV SDGNGMNAWVAWRNRCKGTDVQAWIRGCRL Share Similar Sequence Homologous

In an ideal world, we would be able to accurately predict protein structure from the sequence only! Because of the myriad possible configurations of a protein chain – This goal can’t reliably be achieved, yet. Knowledge based prediction vs. Simulation based on physical forces. Here we will only concern ourselves with knowledge-based methods, although we might use simulation in order to optimize our models. Structure prediction

MNIFEMLRID EGLRLKIYKD TEGYYTIGIG HLLTKSPSLN AAKSELDKAI GRNCNGVITK DEAEKLFNQD VDAAVRGILR NAKLKPVYDS LDAVRRCALI NMVFQMGETG VAGFTNSLRM LQQKRWDEAA VNLAKSRWYN QTPNRAKRVI TTFRTGTWDA YKNL Can we predict protein structures ? ab initio folding simulation: not yet... Rosetta approach: neither... Fold recognition (threading): Often works, but... ???

obtain sequence (target) fold assignment comparative modeling ab initio modeling build, assess model Approaches to predicting protein structures

Homology Modelling of Proteins Definition: Prediction of three dimensional structure of a target protein from the amino acid sequence (primary structure) of a homologous (template) protein for which an X-ray or NMR structure is available. Why a Model: A Model is desirable when either X-ray crystallography or NMR spectroscopy cannot determine the structure of a protein in time or at all. The built model provides a wealth of information of how the protein functions with information at residue property level. This information can than be used for mutational studies or for drug design.

Homology modeling = Comparative protein modeling = Knowledge-based modeling Idea: Extrapolation of the structure for a new (target) sequence from the known 3D-structures of related family members (templates).

Homology models have RMSDs less than 2Å more than 70% of the time. Homology models can be very smart!

identity Number of residues aligned Percentage sequence identity/similarity (B.Rost, Columbia, NewYork) Sequence identity implies structural similarity Don’t know region..... Sequence similarity implies structural similarity?

Step 1 in Homology Modeling - Fold Identification Aim: To find a template or templates structures from protein data base Improved Multiple sequence alignment methods improves sensitivity - remote homologs PSIBLAST, CLUSTAL pairwise sequence alignment - finds high homology sequences BLAST

Comparative Modeling Known Structures (Templates) Target Sequence Template Selection Alignment Template - Target Structure modeling Structure Evaluation & Assessment Homology Model(s) Protein Data Bank PDB  Database of templates Separate into single chains Remove bad structures (models) Create BLAST database

Model Building from template Multiple templates Protein Fold Core conserved regions Variable Loop regions Side chains Calculate the framework from average of all template structures Generate one model for each template and evaluate

I. Manual Modeling [ ]

averaging core template backbone atoms (weighted by local sequence similarity with the target sequence) Leave non-conserved regions (loops) for later …. a) Build conserved core framework II. Template based fragment assembly

Dressing up the Core Model Core Model-Rigid Body Assembly Add Side chains Add loops End Game in protein folding - Molecular dynamics of all atoms in explicit solvent

use the “spare part” algorithm to find compatible fragments in a Loop-Database “ab-initio” rebuilding of loops (Monte Carlo, molecular dynamics, genetic algorithms, etc.) b) Loop modeling II. Template based fragment assembly

Loop Builders Mini protein folding problem- 3 to 10 residues longer in membrane proteins Ab Initio methods - generates various random conformations of loops and score Compare the loop sequence string to DB and get hits and evaluate. Some Homology modeling methods have less number of loops to be added because of extensive multiple sequence alignment of profiles Loops result from substitutions, insertions and deletions in the same family

Using database of loops which appear in known structures. The loops could be catagorised by their length or sequence Ab initio methods - without any prior knowledge. This is done by empirical scoring functions that check large number of conformations and evaluates each of them. Construction of loops might be done by:

c) Side Chain placement Find the most probable side chain conformation, using homologues structures back-bone dependent rotamer libraries energetic and packing criteria II. Template based fragment assembly

modeling will produce unfavorable contacts and bonds  idealization of local bond and angle geometry extensive energy minimization will move coordinates away  keep it to a minimum SwissModel is using GROMOS 96 force field for a steepest descent d) Energy minimization II. Template based fragment assembly

d) Energy minimization II. Template based fragment assembly

Homology Modeling Programs Modeller ( Swiss-Model ( Whatif (

Swiss-Model Method: Knowledge-based approach. Requirements: At least one known 3D-structure of a related protein. Good quality sequence alignements. Procedures: Superposition of related 3D-structures. Generation of a multiple a alignement. Generation of a framework for the new sequence. Rebuild lacking loops. Complete and correct backbone. Correct and rebuild side chains. Verify model structure quality and check packing. Refine structure by energy minimisation and molecular dynamics.

Model Confidence Factors The Model B-factors are determined as follows: The number of template structures used for model building. The deviation of the model from the template structures. The Distance trap value used for framework building. The Model B-factor is computed as: 85.0 * (1/ # selected template str.) * (Distance trap / 2.5) and 99.9 for all atoms added during loop and side-chain building

Verifying the Model PROCHECK WHAT IF PROSA II VERIFY 3D, Profile3D

Errors in Models !!! Incorrect template selection Incorrect alignments Errors in positioning of sidechains and loops

General Structure Prediction Scheme Any given protein sequence Structure selection Check sequence identity with proteins with known structure Homology Modeling > 35% Fold Recognition ab initio Folding < 35% Structure refinement Final Structure Structure selection

Baker and Sali (2000)

EVA Evaluation of Automatic protein structure prediction [ Burkhard Rost, Andrej Sali, ] CASP Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction 3D - Crunch Very Large Scale Protein Modelling Project Model Accuracy Evaluation

Several web pages for homology modeling COMPOSER – felix.bioccam.ac.uksoft-base.html MODELLER – guitar.rockefeller.edu/modeller/modeller.html WHAT IF – SWISS-MODEL –