Types of proteinprotein interactions PPI

1
(No Transcript)
2
Types of protein-protein interactions (PPI)
Non-obligate PPI
Obligate PPI the protomers are not found as
stable structures on their own in vivo
Non-obligate homodimer Sperm lysin
Obligate homodimer P22 Arc repressor DNA-binding
Obligate heterodimer Human cathepsin D
Non-obligate heterodimer RhoA and RhoGAP
signalling complex
3
Types of protein-protein interactions (PPI)
Non-obligate PPI
Obligate PPI usually permanent the protomers
are not found as stable structures on their own
in vivo
Permanent (many enzyme-inhibitor
complexes) dissociation constant KdAB / AB
10-7 10-13 M
Transient
Weak (electron transport complexes) Kd mM-?M
Non-obligate transient homodimer, Sperm lysin
(interaction is broken and formed continuously)
Intermediate (antibody-antigen,
TCR-MHC-peptide, signal transduction PPI), Kd
?M-nM
Strong (require a molecular trigger to shift the
oligomeric equilibrium) Kd nM-fM
Obligate heterodimer Human cathepsin D
Non-obligate permanent heterodimer Thrombin and
rodniin inhibitor
Bovine G protein dissociates into G? and G??
subunits upon GTP, but forms a stable trimer upon
GDP
4
Is the whole protein surface available for
interaction with other proteins, or are specific
sites pre-assigned according to their biophysical
and structural character? Neuvirth et al., JMB,
2004

• 3D structures of protein-protein complexes have
  been extensively analysed, including
• Amino acid composition of different protein
  regions.
• Residue-residue preferences at the binding site.
• Hydrophobicity of the protein surface, binding
  site, and core region.
• Electrostatic complimentarity of binding
  regions.
• Intermolecular hydrogen binding and salt bridges
  in protein complexes.
• Geometry (size, shape) and sterical
  complementarity of binding sites.
• Most of studies devote to all complexes despite
  their nature.
• Recent studies attempt to treat complexes
  according their nature.

5
Sequence conservancy

• Yes
• Caffrey et al., Protein Sci, 2004
• 64 protein-protein interfaces (different nature)
  
• conservation scores derived from two multiple
  sequence alignment types, one of close
  homologs/orthologs and one of diverse
  homologs/paralogs.
• found that the interface is slightly more
  conserved than the rest of the protein surface
• found that obligate interfaces differ from
  transient interfaces in that the former have
  buried interface residues that are more conserved
  than partially buried interface residues.
• Jones Thornton, JMB, 1997
• Lo Conte et al., JMB, 1999
• Ofran Rost, JMB, 2003

• No
• Glaser et al., Proteins, 2001
• 621 protein-protein interfaces (different
  nature)
• compared residue composition at the interprotein
  interfaces to that at intraprotein contacts and
  at whole genomes
• found that protein interfaces are not
  significantly different than other regions of the
  protein
• Keskin et al., Protein Sci., 1998

6
Statistical analysis of predominantly transient
protein-protein interfaces Ansari Helms
(Proteins, 2005, online in advance of print)
Figure 2. Amino-acid composition of
protein-protein interfaces (our data) compared to
the general composition in the SwissProt
database. Our data were retrieved from a distance
criterion of 5 Å between two interacting chains
of 170 interfaces (antigen-antibody complexes
have been excluded).
The average distribution is 30.4 hydrophobic,
32.8 hydrophilic uncharged, and 36.8 charged
residues. In contrast to other studies on
protein-protein interfaces, the charged residues
are the largest fraction Tsai et al., Crit Rev
Biochem Mol Biol 1996 Lijnzaad Argos Proteins
1997 Zhou Shan Proteins 2001. The higher
occurrence of methionine, phenylalanine,
tryptophan, cysteine, and histidine compared to
the Swiss-Prot distribution could be a
statistical balance of the underrepresentation of
hydrophobic amino acids such as alanine, valine,
leucine, and isoleucine.
7
The results are sensitive to the dataset Ansari
Helms (2005) didnt find correlation of
residue-residue preferences for transient
hetero-complexes obtained on their manually
generated dataset with those obtained on
automatically generated datasets by Ofran Rost
(JMB, 2003) and Glaser et al. (2001) even the
former one contained only transient complexes,
but the second one 217 interfaces of heterodimers
both obligate and transient.
Figure 3. Amino-acid pairing preferences matrix
of transient protein-protein interfaces. The
scores are the normalized pairing frequencies of
two residues that occur on the protein-protein
interfaces of predominantly transient complexes.
(Ansari Helms).
Figure 2. Residueresidue preferences. (F)
transient hetero-oligomers (hetero-complexes). A
red square indicates that the interaction occurs
more frequently than expected a blue square
indicates that it occurs less frequently than
expected. (Ofran Rost).
8
No single parameter absolutely differentiates
interfaces from other surface patches. It was
shown by Jones Thornton (1997, 28 homo-dimers,
20 hetero-dimers) and confirmed by others on more
extensive datasets. Thus all existing method
attempting to predict protein-protein binding
sites combine a number of physical-chemical
properties. The last published method (todays
paper) uses the similar strategy. Most of the
methods dont treat interfaces depending on their
nature. The considered one does. Since the
methods are sensitive to the datasets they are
trained and tested on, it is difficult to compare
them. Todays paper provides a detailed
comparison with two other similar methods, recent
one ( Neuvirth et al., JMB, 2004) and mostly
sited one (Jones Thornton, JMB, 1997). Most of
the methods are not available in any form. The
considered today is available as a server.
9
Dataset

• The following complexes have been excluded
• NMR structures
• mutant complexes
• structures with resolution gt3Å
• proteins sharing gt20 sequence homology
• with interfaces made up of more than one separate
  chain
• with broken interfaces
• If no evidence in literature that the complex
  occur naturally and was stable as a dimer
• For homodimers subunit with the largest binding
  site was retained.
• 180 proteins (from 149 complexes) survived
• 36 involved in enzyme-inhibitor interactions
• 27 involved in hetero-obligate interactions,
• 87 involved in homo-obligate interaction
• 30 involved in non-enzyme-inhibitor transient
  interactions (NEIT).

10

• Atom was defined as part of interface if it lost
  gt99 of its accessible surface area upon complex
  formation.
• Surface patches generation (procedure similar to
  described in Jones Thornton, 1997) (one patch
  per surface atom) of the sizes between 6 and 13
  of the smallest protein in the complex.
• Seven parameters describing a patch
• Surface shape index (values from -1 to 1)
  defined from the principal curvatures.
• Surface shape curvedness defined from the
  principal curvatures.
• Electrostatic potential calculated by the Delphi
  4 package (Rocchia et al., 2002)
• Sequence conservation score calculated by the
  Scorecons program (Valdar, 2002) for a residue
  surface patch vertex corresponds to.
• Hydrophobicity calculated using the Fauchere
  Pliska (1983) scale for a residue surface patch
  vertex corresponds to.
• Residue interface propensity calculated from an
  authors dataset for each of 20 amino acids as a
  fraction of the solvent excluded surfaces (SES)
  that each aa contributes to the interface
  compared to the fractions of the SES that each aa
  contributes to the whole protein surface.
• ASA of each atom.
• Each surface vertex within a patch was labeled
  with seven parameters, which were normalized
  between 0 and 1, mean and standard deviation of
  each property was calculated across the patch.

11

• SVM has been trained on 14 attributes (7
  parameters for interface and non-interface patch)
  using leave-one-out cross-validation repeated 5
  times.
• Specificity
• number of interface residues in patch / number
  of patch residues
• Sensitivity
• number of interface residues in patch /
• number of interface residues
• A prediction was deemed a success if a patch
  with over 50 specificity and 20 sensitivity was
  ranked in the top three.
• Success rates
• 64 (23/36) for enzyme-inhibitory PPI
• 85 (23/27) for heter-obligomers
• 82 (93/114) for obligate binding sites
• 65 (43/66) for transient binding sites

12
(No Transcript)
13
(No Transcript)
14
Cases when sensitivity values are higher than
specificity values highlight the problem of
choosing an appropriate patch size.
15
Control experiment HIV-1 envelope protein gp120
core complexed with CD4 and a neutralizing human
antibody 17b (1gc1). 17b epitope (1gc1G) is
comprised of four discontinuous
?-strands. Neither gp120 nor any its homolog has
been used in the study of Bradford
Westhead. The best predicted binding patch of
gp120 protein contains 73 residues including 13
from 15 residues of 17b epitope. Sensitivity
86 (13 / 15) Specificity 18 (13 / 73)
17b epitope
16
Comparison with other patch analysis methods
(Jones Thornton, 1997)

• Success rates
• 72 (34/47) on the whole data set
• 79 (15/19) on enzyme-inhibitor
• 33 (2/6) for NEIT interaction types, possibly
  because they made up the minority of the training
  set.

Jones and Thornton achieved a success rate of 64
(30/47) on the 47 test cases. In 41 out of the
47 test cases, patch sizes used by Jones and
Thornton were larger than ours. If we had
increased our patch size then our sensitivity
values would have increased as well but at the
risk of decreasing specificity (not reported by
Jones and Thornton). We also defined our
interfaces more stringently than Jones and
Thornton who included residues that lost more
than 1 Å2 accessible surface area upon complex
formation whereas we only used residues that lost
more than 99 accessible surface area. As a
result, Jones and Thornton generated larger
interfaces than us in most cases thus
exaggerating sensitivity values in patches around
the edge of the interface. Taken together, the
larger patch sizes and interfaces used by Jones
and Thornton would have increased the probability
of an apparently successful outcome. This is
reflected in the numbers of successes one would
expect to obtain by random sampling of patches.
Based on data from their paper, we estimate that
17 of the 30 successes achieved by Jones and
Thornton could have been due to random chance.
This compares to only 15 of the 34 successes that
we attained.
17
Comparison with other patch analysis methods
(Neuvirth et al., 2004)

• Success rates
• 51 (44/87) on the whole data set
• 56 (14/25) on enzyme-inhibitor
• 44 (24/54) for NEIT interaction types, possibly
  because they made up the minority of the training
  set.

Neuvirth et al. achieved a success rate of 62
(33/53) on their unbound data set with the
criterion for success being a predicted patch
with over 50 specificity. However, applying
our own criteria of 50 specificity and 20
sensitivity reduced their success rate to 36
(19/53). On the same set of proteins in their
bound form, our success rate was 53 (28/53).
Neuvirth et al. put even greater emphasis on
specificity than we did to the detriment of their
sensitivity values. We agree with Neuvirth et
al. that specificity is important but a good
prediction should at least cover 20 of the
interface. We believe our method provides the
best balance between specificity and sensitivity
of the two methods. For example, the mean
sensitivity of patches with over 50 specificity
from our method was 51 compared to approximately
20 from the Neuvirth et al. method. The
negative influence of including so many proteins
involved in NEIT interactions was illustrated by
the success rates of each interaction type
separately. This suggested that additional
attributes are needed that would better
distinguish the interface from the rest of the
surface of proteins involved in NEIT interactions.
18
Important properties obtained by the training the
SVM on each property separately repeated ten
times
An MCC of 1 represents perfect training
classification (no false positives or negatives)
whereas 1 represents a complete failure (all
interacting patches classified as non-interacting
patches and vice versa). Attributes based on
interface residue propensity, hydrophobicity and
ASA achieved the highest MCC values. Shape index
and conservation also performed well.
Electrostatic potential seemed to have some
differentiating power even though it has been
used in past studies more successfully to predict
DNA binding sites (Jones et al. 2003). In
general, using the SVM on all attributes gave
better performance than any one attribute alone,
indicating that attributes give complementary
information. The higher MCC value achieved with
curvedness on transient interfaces probably
reflects the number of enzyme-inhibitor
interfaces in the subset. It is common for a
protrusion on the inhibitor surface to bind
inside a cleft on the enzyme surface and these
protrusions and clefts will be highly curved.
Electrostatic potential has almost no
distinguishing power on transient interfaces,
achieving an MCC value of only 0.08?0.01 in
contrast to obligomeric interfaces where it
achieves an MCC value of 0.27?0.05.
A higher MCC value (0.72?0.03) was achieved with
training on obligomeric interfaces using all
attributes than with transient interfaces
(0.63?0.04) suggesting that obligomeric
interfaces contain stronger signals that
distinguish them from the rest of the protein
surface than transient interfaces.