Transcript Protein-protein interactions.

Protein structure prediction.
Protein folds.
• Fold definition: two folds are similar if they
have a similar arrangement of SSEs
(architecture) and connectivity (topology).
Sometimes a few SSEs may be missing.
• Fold classification: structural similarity
between folds is searched using structurestructure comparison algorithms.
Protein structure prediction flowchart
Protein
sequence
Does
sequence
align with a
protein of
known
structure?
Database
similarity
search
No
Protein
family
analysis
Yes
Predicted threedimensional
structural model
Three-dimensional
comparative
modeling
Yes
Threedimensional
structural analysis
in laboratory
No
Is there a
predicted
structure?
Yes
Relationship
to known
structure?
No
Structural
analysis
From D.W.Mount
Protein structure prediction.
Prediction of three-dimensional structure of a protein from its
sequence. Different approaches:
- Homology modeling (query protein has a very close homolog
in the structure database).
- Fold recognition (query protein can be mapped to template
protein with the existing fold).
- Ab initio prediction (query protein has a new fold).
Homology modeling.
Aims to produce protein models with accuracy
close to experimental and is used for:
- Protein structure prediction
- Drug design
- Prediction of functionally important sites (active
or binding sites)
Steps of homology modeling.
1.
2.
3.
4.
5.
Template recognition & initial alignment.
Backbone generation.
Loop modeling.
Side-chain modeling.
Model optimization.
1. Template recognition.
Recognition of similarity between the target and template.
Target – protein with unknown structure.
Template – protein with known structure.
Main difficulty – deciding which template to pick, multiple
choices/template structures.
Template structure can be found by searching for structures
in PDB using pairwise sequence alignment methods.
Two zones of protein structure prediction.
Sequence identity
100
Homology modeling zone
50
Fold recognition zone
50
100
150
200
Alignment length
2. Backbone generation.
If alignment between target and template is ready,
copy the backbone coordinates of those
template residues that are aligned.
If two aligned residues are the same, copy their
side chain coordinates as well.
3. Insertions and deletions.
insertion
AHYATPTTT
AH---TPSS
deletion
Occur mostly between secondary structures, in the loop
regions. Loop conformations – difficult to predict.
Approaches to loop modeling:
- Knowledge-based: search the PDB for loops with known
structures
- Energy-based: an energy function is used to evaluate the
quality of a loop. Energy minimization or Monte Carlo.
4. Side chain modeling.
Side chain conformations – rotamers. In similar proteins side chains have similar conformations.
If % identity is high - side chain conformations can be copied
from template to target. If % identity is not very high modeling of side chains using libraries of rotamers and
different rotamers are scored with energy functions.
Problem: side chain configurations depend on backbone
conformation which is predicted, not real
E2
E3
E1
E = min(E1, E2, E3)
5. Model optimization.
Energy optimization of entire structure.
Since conformation of backbone depends on
conformations of side chains and vice versa iteration approach:
Predict rotamers
Shift in backbone
Classwork: Homology modeling.
-
Go to NCBI Entrez, search for gi461699
Do Blast search against PDB
Repeat the same for gi60494508
Compare the results
Fold recognition.
Unsolved problem: direct prediction of protein structure from
the physico-chemical principles.
Solved problem: to recognize, which of known folds are
similar to the fold of unknown protein.
Fold recognition is based on observations/assumptions:
- The overall number of different protein folds is limited
(1000-3000 folds)
- The native protein structure is in its ground state (minimum
energy)
Fold recognition.
Goal: to find protein with known structure which best
matches a given sequence.
Since similarity between target and the closest
template is not high, pairwise sequence alignment
methods fail.
Solution: threading – sequence-structure alignment
method.
Threading – method for structure
prediction.
Sequence-structure alignment, target sequence is
compared to all structural templates from the
database.
Requires:
- Alignment method (dynamic programming, Monte
Carlo,…)
- Scoring function, which yields relative score for
each alternative alignment
Protein structure prediction: target sequence is
compared to structures using sequencestructure alignment
Structural templates
Score1
Score2
Target sequence
Concept of threading: D. Jones et al, 1993
Score3
Protein structure prediction: target sequence is
compared to structures using sequencestructure alignment
Structural templates
Score1
Score2
Target sequence
Score3>Score2>Score1
Score3
Structural
model of target
Scoring function for threading.
• Contact-based scoring function
depends on amino acid types of two
residues and distance between
them.
• Sequence-sequence alignment
scoring function does not depend on
the distance between two residues.
• If distance between two nonadjacent residues in the template is
less than 8 Å, these residues make
a contact.
Scoring function for threading.
Ala
Trp
Tyr
Ile
S
N
 w(a , a
i , j 1
i
j
); S  w( Ala, Tyr)  w( Ile, Trp)
“w” is calculated from the frequency of amino acid contacts in
PDB; ai – amino acid type of target sequence aligned with the
position “i” of the template; N- number of contacts
Classwork: calculate the score for target sequence
“ATPIIGGLPY” aligned to template structure which
is defined by the contact matrix.
A
T
1
1
2
3
4
*
5
6
*
7
8
9
10
*
Y
2
3
I
G
*
*
6
7
*
*
*
8
*
9
10
T
P
Y
I
G
L
-0.2
-0.1
0
-0.1
0.5
-0.2
0.2
0.3
-0.1
-0.2
-0.3
0.1
0
-0.2
-0.4
-0.1
0.1
-0.2
-0.4
-0.2
-0.1
-0.2
0.3
0.2
0.4
0.4
0.2
*
4
5
P
A
*
*
L
0.3
Optimize the Sum of
Residue-Residue
Contact Potentials ...
…. by a Monte Carlo
Alignment Algorithm
Evaluation of quality of structural model
• Correct bond length and bond angles
>> 3.8 Angstroms
• Correct placement of functionally important sites
• Prediction of global topology, not partial alignment
(minimum number of gaps)
Success and limitations of structure prediction
Limitations:
Success:
•
•
Accuracy scores almost doubled
from CASP1 to CASP6, might be
because of database size
Models of small targets are very
accurate
•
•
•
Models of large and remotely
related proteins are not very
accurate
Domain boundaries are difficult to
define
Models often do not provide details
for functional annotation
Adapted from Kryshtafovych et al 2005
GenThreader
http://bioinf.cs.ucl.ac.uk/psipred.
1. Predicts secondary structures for target
sequence.
2. Makes sequence profiles (PSSMs) for
each template sequence.
3. Uses threading scoring function to find
the best matching profile.
Protein-protein interactions.
Common properties of protein-protein interactions.
rim
• Majority of protein complexes have a buried
surface area ~1600±400 Ǻ^2 (“standard size”
patch).
• Complexes of “standard size” do not involve
large conformational changes while large
complexes do.
core
Top molecule
• Protein recognition site consists of a completely
buried core and a partially accessible rim.
• Trp and Tyr are abundant in the core, but Ser
and Thr, Lys and Glu are particularly disfavored.
Bottom molecule
Different types of protein-protein interactions.
• Permanent and transient.
• External are between different chains; internal are within
the same chain.
• Homo- and hetero-oligomers depending on the similarity
between interacting subunits.
• Interface type can be predicted from amino acid
composition (Ofran and Rost 2003).
Experimental methods
Verification of experimental protein-protein interactions.
• Protein localization method.
• Expression profile reliability method.
• Paralogous verification method.
Protein localization method.
Sprinzak, Sattath, Margalit, J Mol Biol,
2003
A – A3: Y2H
B: physical methods
C: genetics
E: immunological
True positives:
- Proteins which are localized in the
same cellular compartment
- Proteins with a common cellular role
Expression profile reliability method.
Deane, C. M. (2002)
Mol. Cell. Proteomics 1: 349-356
Paralogous verification method.
PVM method is based on
observation that if two proteins
interact, their paralogs would
interact. Calculates the number
of interactions between two
families of paralogous proteins.
Deane, C. M. (2002)
Mol. Cell. Proteomics 1: 349-356
Interaction databases
• Experiment (E)
• Structure detail (S)
• Predicted
– Physical (P)
– Functional (F)
• Curated (C)
• Homology
modeling (H)
• *IMEx consortium
Protein interaction databases
• Protein-protein interaction databases
• Domain-domain interaction databases
DIP database
• Documents proteinprotein interactions from
experiment
– Y2H, protein microarrays,
TAP/MS, PDB
• 55,733 interactions
between 19,053 proteins
from 110 organisms.
Organisms
# proteins # interactions
Fruit fly
7052
20,988
H. pylori
710
1425
Human
916
1407
E. coli
1831
7408
C. elegans
2638
4030
Yeast
4921
18,225
Others
985
401
DIP database
Duan et al., Mol Cell Proteomics, 2002
• Assess quality
– Via proteins: PVM, EPR
– Via domains: DPV
• Search by BLAST or
identifiers / text
BIND database
Alfarano et al., Nucleic Acids Res, 2005
• Records experimental
interaction data
• 83,517 protein-protein
interactions
• 204,468 total interactions
• Includes small molecules,
NAs, complexes
Classwork.
• Go to DIP webpage (http://dip.doembi.ucla.edu)
• Retrieve all interactions for cytochrome C,
tubulin, RNA-polymerase from yeast
• How many of them are confirmed by
several experimental methods?
Protein interaction databases
• Protein-protein interaction databases
• Domain-domain interaction databases
InterDom database
Ng et al., Nucleic Acids Res, 2003
• Predicts domain
interactions (~30000)
from PPIs
• Data sources:
–
–
–
–
Domain fusions
PPI from DIP
Protein complexes
Literature
• Scores interactions
Pibase database
• Records domain interactions from PDB and PQS
• Domains defined with SCOP and CATH
• All inter-domain and inter-chain distances within 6 Ǻ are
considered interacting domains
• From interacting domain pairs, create list of interfaces
with buried solvent accessible area > 300 Ǻ2
Classwork.
• Go to Pibase website
http://salilab.org/pibase
• Select largest structural complexes, 1k73,
1i6h
• Compare two complexes in terms of the
number of interacting domains,
#interactions per node
NCBI CBM database
Shoemaker et al., Protein Sci, 2006.
• CBM – database of interacting structural domains exhibiting
Conserved Binding Modes
• To retrieve interactions:
– Record interactions
– Use VAST structural
alignments to compare
binding surfaces
– Study recurring domaindomain interactions
Definition of CBM
• Interacting domain pair – if at least 5
residue-residue contacts between
domains (contacts – distance of less
than 8 Ǻ)
• Structure-structure alignments
between all proteins corresponding to
a given pair of interacting domains
• Clustering of interface similarity, those
with >50% equivalently aligned
positions are clustered together
• Clusters with more than 2 entries
define conserved binding mode.
Number of interacting pairs and binding modes
• 833 conserved interaction types
• 1,798 total domain interaction types
• Up to 24 CBMs per interaction type
CBM
Structures
Species
1
154
Jawed vertebrates
2
112
Jawed vertebrates
3
17
Clam,earthworm
4
4
lamprey
5
4
V.stercoraria
6
2
Rice,soybeans
7
2
human
8
2
lamprey
• Classify complicated domain
pairs by CBMs
• Globin example:
– 630 pairs
– 2 CBMs account for majority
Classwork.
• Retrieve structures 1GY3, 1E9H, 1OL2
• Examine all interactions within and
between chains/domains.
• How many CBMs do you find?