Prediction of protein structure, homology Modeling

Prediction of protein structure

2. aim
Structure prediction tries to build models of 3D
structures of proteins that could be useful for
understanding structure-function relationships.

3. Genbank/EMBL 105.000.000
Uniprot5.200.000
PDB 47.000

5. The protein folding problem
The information for 3D structures is coded in the
protein sequence
Proteins fold in their native structure in seconds
Native structures are both thermodynamically
stables and kinetically available

6. AVVTW...GTTWVRAVVTW...GTTWVR
ab-initio prediction
Prediction from sequence using first principles

7. Ab-initio prediction
“In theory”, we should be able to build native
structures from first principles using sequence
information and molecular dynamics
simulations: “Ab-initio prediction of structure”
Simulaciones de 1 s de “folding” de una proteína modelo
(Duan-Kollman: Science, 277, 1793, 1998).
Simulaciones de folding reversible de péptidos (20-200 ns)
(Daura et al., Angew. Chem., 38, 236, 1999).
Simulaciones distribuidas de folding de Villin (36-residues)
(Zagrovic et al., JMB, 323, 927, 2002).

8. ... the bad news ...
It is not possible to span simulations to the
“seconds” range
Simulations are limited to small systems and fast
folding/unfolding events in known structures
steered dynamics
biased molecular dynamics
Simplified systems

9. typical shortcuts
Reduce conformational space
1,2 atoms per residue
fixed lattices
Statistic force-fields obtained from known structures
Average distances between residues
Interactions
Use building blocks: 3-9 residues from PDB
structures

10. Some protein from ESome protein from E.coli.coli
predicted at 7.6 Åpredicted at 7.6 Å
(CASP3, H.Scheraga)(CASP3, H.Scheraga)
Results from ab-initio
Average error 5 Å - 10 Average error 5 Å - 10
ÅÅ
Function cannot be Function cannot be
predictedpredicted
Long simulationsLong simulations

11. comparative modelling
The most efficient way to predict protein
structure is to compare with known 3D
structures

12. Protein folds

13. Basic concept
In a given protein 3D structure is a more
conserved characteristic than sequence
Some aminoacids are “equivalent” to each other
Evolutionary pressure allows only aminoacids
substitutions that keep 3D structure largely
unaltered
Two proteins of “similar” sequences must have
the “same” 3D structure

14. Possible scenarios
1. Homology can be recognized using sequence comparison tools or
protein family databases (blast, clustal, pfam,...).
Structural and functional predictions are feasible
2. Homology exist but cannot be recognized easily (psi-blast,
threading)
Low resolution fold predictions are possible. No functional
information.
3. No homology
1D predictions. Sequence motifs. Limited functional prediction.
Ab-initio prediction

15. fold prediction

16. 3D struc. prediction

17. 1D prediction
Prediction is based on averaging aminoacid
properties
AGGCFHIKLAAGIHLLVILVVKLGFSTRDEEASS
Average over a
window

19. 1D prediction. Properties
Secondary structure propensitites
Hydrophobicity (transmembrane)
Accesibility
...

20. Aminoacido P() P() P(turn)
Ala 1.29 0.9 0.78
Cys 1.11 0.74 0.8
Leu 1.3 1.02 0.59
Met 1.47 0.97 0.39
Glu 1.44 0.75 1
Gln 1.27 0.8 0.97
His 1.22 1.08 0.69
Lys 1.23 0.77 0.96
Val 0.91 1.49 0.47
Ile 0.97 1.45 0.51
Phe 1.07 1.32 0.58
Tyr 0.72 1.25 1.05
Trp 0.99 1.14 0.75
Thr 0.82 1.21 1.03
Gly 0.56 0.92 1.64
Ser 0.82 0.95 1.33
Asp 1.04 0.72 1.41
Asn 0.9 0.76 1.23
Pro 0.52 0.64 1.91
Arg 0.96 0.99 0.88
Propensities Chou-Fasman
Biochemistry 17, 4277 1978


turn

21. Some programs (www.expasy.org)
BCM PSSP - Baylor College of Medicine
Prof - Cascaded Multiple Classifiers for Secondary Structure Prediction
GOR I (Garnier et al, 1978) [At PBIL or at SBDS]
GOR II (Gibrat et al, 1987)
GOR IV (Garnier et al, 1996)
HNN - Hierarchical Neural Network method (Guermeur, 1997)
Jpred - A consensus method for protein secondary structure prediction
at University of Dundee
nnPredict - University of California at San Francisco (UCSF)
PredictProtein - PHDsec, PHDacc, PHDhtm, PHDtopology, PHDthreader,
MaxHom, EvalSec from Columbia University
PSA - BioMolecular Engineering Research Center (BMERC) / Boston
PSIpred - Various protein structure prediction methods at Brunel
University
SOPM (Geourjon and Deléage, 1994)
SOPMA (Geourjon and Deléage, 1995)
AGADIR - An algorithm to predict the helical content of peptides

22. 1D Prediction
Original methods: 1 sequence and uniform
parameters (25-30%)
Original improvements: Parameters specific
from protein classes
Present methods use sequence profiles obtained
from multiple alignments and neural networks
to extract parameters (70-75%, 98% for
transmembrane helix)

24. Methods for remote homology
Homology can be recognized using PSI-Blast
Fold prediction is possible using threading
methods
Acurate 3D prediction is not possible: No
structure-function relationship can be inferred
from models

25. Threading
Unknown sequence is “folded” in a number of
known structures
Scoring functions evaluate the fitting between
sequence and structure according to statistical
functions and sequence comparison

26. ..........
10.510.5 5.2>> ..........
SELECTED HITSELECTED HIT

27. ATTWV....PRKSCTATTWV....PRKSCT SequenceSequence
HHHHH....CCBBBBHHHHH....CCBBBB Pred. Sec. Struc.Pred. Sec. Struc.
eeebb....eeebebeeebb....eeebeb Pred. accesibilityPred. accesibility
..........
SequenceSequence GGTV....ATTW ........... ATTVL....FFRKGGTV....ATTW ........... ATTVL....FFRK
Obs SS Obs SS BBBB....CCHH ........... HHHB.....CBCB BBBB....CCHH ........... HHHB.....CBCB
Obs Acc. Obs Acc. EEBE.....BBEB ........... BBEBB....EBBEEEBE.....BBEB ........... BBEBB....EBBE

28. Threading accurancyThreading accurancy
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
%

A
C
I
E
R
T
O
S
5 10 15 20 25
% IDENTIDAD SECUENCIAS

31. Comparative modelling
Good for homology >30%
Accurancy is very high for homology > 60%
Reminder
The model must be USEFUL
Only the “interesting” regions of the protein need
to be modelled

32. Expected accurancy
Strongly dependent on the quality of the sequence
alignment
Strongly dependent on the identity with “template”
structures. Very good structures if identity > 60-70%.
Quality of the model is better in the backbone than
side chains
Quality of the model is better in conserved regions

34. Quality test
No energy differences between a correct or
wrong model
The structure must by “chemically correct” to
use it in quantitative predictions

35. Analysis software
PROCHECK
WHATCHECK
Suite Biotech
PROSA

39. Prediction software
SwissModel (automatic)
http://www.expasy.org/swissmod/
SwissModel Repository
http://swissmodel.expasy.org/repository/
3D-JIGSAW (M.Stenberg)
http://www.bmm.icnet.uk/servers/3djigsaw/
Modeller (A.Sali)
http://salilab.org/modeller/modeller.html
MODBASE (A. Sali)
http://alto.compbio.ucsf.edu/modbase-cgi/index.cgi

40. Final test
The model must justify experimental data (i.e.
differences between unknown sequence and
templates) and be useful to understand function.