Protein Tertiary Structure Prediction Structural Bioinformatics
Primary: amino acid linear sequence. Secondary: -helices, β-sheets and loops. Tertiary: the 3D shape of the fully folded polypeptide chain The Different levels of Protein Structure
3 PDB: Protein Data Bank DataBase of molecular structures : Protein, Nucleic Acids (DNA and RNA), Structures solved by X-ray crystallography NMR Electron microscopy
4 RCSB PDB – Protein Data Bank
How can we view the protein structure ? Download the coordinates of the structure from the PDB Launch a 3D viewer program For example we will use the program Pymol The program can be downloaded freely from the Pymol homepage Upload the coordinates to the viewer
Pymol example Launch Pymol Open file “1aqb” (PDB coordinate file) Display sequence Hide everything Show main chain / hide main chain Show cartoon Color by ss Color red Color green, resi 1:40 Help
Predicting 3D Structure –Comparative modeling (homology) Based on structural homology –Fold recognition (threading) Outstanding difficult problem Based on sequence homology
Comparative Modeling Similar sequences suggests similar structure
Sequence and Structure alignments of two Retinol Binding Protein
Structure Alignments The outputs of a structural alignment are a superposition of the atomic coordinates and a minimal Root Mean Square Distance (RMSD) between the structures. The RMSD of two aligned structures indicates their divergence from one another. Low values of RMSD mean similar structures There are many different algorithms for structural Alignment.
Dali (Distance mAtrix aLIgnment) DALI offers pairwise alignments of protein structures. The algorithm uses the three- dimensional coordinates of each protein to calculate distance matrices comparing residues. See Holm L and Sander C (1993) J. Mol. Biol. 233: SALIGN
Comparative Modeling Builds a protein structure model based on its alignment to one or more related protein structures in the database Similar sequence suggests similar structure
Comparative Modeling Accuracy of the comparative model is related to the sequence identity on which it is based >50% sequence identity = high accuracy 30%-50% sequence identity= 90% modeled <30% sequence identity =low accuracy (many errors)
Homology Threshold for Different Alignment Lengths Alignment length (L) Homology Threshold (t) A sequence alignment between two proteins is considered to imply structural homology if the sequence identity is equal to or above the homology threshold t in a sequence region of a given length L. The threshold values t(L) are derived from PDB
Comparative Modeling Similarity particularly high in core –Alpha helices and beta sheets preserved –Even near-identical sequences vary in loops
Comparative Modeling Methods MODELLER (Sali –Rockefeller/UCSF) SCWRL (Dunbrack- UCSF ) SWISS-MODEL
Comparative Modeling Modeling of a sequence based on known structures Consist of four major steps : 1.Finding a known structure(s) related to the sequence to be modeled (template), using sequence comparison methods such as PSI-BLAST 2. Aligning sequence with the templates 3. Building a model 4. Assessing the model
Fold Recognition
HemoglobinTIM Protein Folds: sequential and spatial arrangement of secondary structures
Similar folds usually mean similar function Homeodomain Transcription factors
The same fold can have multiple functions Rossmann TIM barrel 12 functions 31 functions
Fold Recognition Methods of protein fold recognition attempt to detect similarities between protein 3D structure that have no significant sequence similarity. Search for folds that are compatible with a particular sequence. "the turn the protein folding problem on it's head” rather than predicting how a sequence will fold, they predict how well a fold will fit a sequence
Basic steps in Fold Recognition : Compare sequence against a Library of all known Protein Folds (finite number) Query sequence MTYGFRIPLNCERWGHKLSTVILKRP... Goal: find to what folding template the sequence fits best There are different ways to evaluate sequence-structure fit
MAHFPGFGQSLLFGYPVYVFGD... Potential fold... 1)... 56)... n) There are different ways to evaluate sequence-structure fit
Programs for fold recognition TOPITS (Rost 1995) GenTHREADER (Jones 1999) SAMT02 (UCSC HMM) 3D-PSSM
Ab Initio Modeling Compute molecular structure from laws of physics and chemistry alone Theoretically Ideal solution Practically nearly impossible WHY ? –Exceptionally complex calculations –Biophysics understanding incomplete
Ab Initio Methods Rosetta (Bakers lab, Seattle) Undertaker (Karplus, UCSC)
CASP - Critical Assessment of Structure Prediction Competition among different groups for resolving the 3D structure of proteins that are about to be solved experimentally. Current state - –ab-initio - the worst, but greatly improved in the last years. –Modeling - performs very well when homologous sequences with known structures exist. –Fold recognition - performs well.
What can you do? FOLDIT Solve Puzzles for Science A computer game to fold proteins
What’s Next Predicting function from structure
Structural Genomics : a large scale structure determination project designed to cover all representative protein structures Zarembinski, et al., Proc.Nat.Acad.Sci.USA, 99:15189 (1998) ATP binding domain of protein MJ0577
As a result of the Structure Genomic initiative many structures of proteins with unknown function will be solved Wanted ! Automated methods to predict function from the protein structures resulting from the structural genomic project.
Approaches for predicting function from structure ConSurf - Mapping the evolution conservation on the protein structure
Approaches for predicting function from structure PFPlus – Identifying positive electrostatic patches on the protein structure
Approaches for predicting function from structure SHARP2 – Identifying positive electrostatic patches on the protein structure
Machine learning approach for predicting function from structure Find the common properties of a protein family (or any group of proteins of interest) which are unique to the group and different from all the other proteins. Generate a model for the group and predict new members of the family which have similar properties.
Knowledge Based Approach Generate a dataset of proteins with a common function (DNA binding protein) Generate a control dataset Calculate the different properties which are characteristic of the protein family you are interested for all the proteins in the data (DNA binding proteins and the non-DNA binding proteins Represent each protein in a set by a vector of calculated features and build a statistical model to split the groups Basic Steps 1. Building a Model
Calculate the properties for a new protein And represent them in a vector Predict whether the tested protein belongs to the family Basic Steps 2. Predicting the function of a new protein
TEST CASE Y14 – A protein sequence translated from an ORF (Open Reading Frame) Obtained from the Drosophila complete Genome >Y14 PQRSVGWILFVTSIHEEAQEDEIQEKFCDYGEIKNIHL NLDRRTGFSKGYALVEYETHKQALAAKEALNGAEIM GQTIQVDWCFVKG G
Support Vector Machine (SVM) To find a hyperplane that maximally separates the RNA-binding from non-RNA binding into two classes Input spaceFeature space Kernel function ? new protein structure RNA binding Non-NA binding =[x1, x2, x3…] =[y1, y2,y3…]
>Y14 PQRSVGWILFVTSIHEEAQEDEIQEKFCDYGEIKNI HLNLDRRTGFSKGYALVEYETHKQALAAKEALN GAEIMGQTIQVDWCFVKG G Y14 DOES NOT BIND RNA