Transcript Document

Protein Sectors: Evolutionary
Units of Three-Dimensional
Structure
Najeeb Halabi, Olivier Rivoire, Stanislas Leibler, and Rama Ranganthan
Cell 138, 774-786, August 21, 2009
Journal Club
Yizhou Yin
Sep 23, 2009
Sequence Conservation
“…sequence conservation – the degree to which the frequency
of amino acids at a given position deviates from random
expectation in a well sampled multiple sequence alignment of
the protein family...”
evolution
sequence
structure
Evolutionary relationship
sequence
conservation
Structural/functional
importance
property/function
Hypothesis
-However, in the 3-dimensional structure of protein, the large
amount of interactions between amino acid residues are also
fundamental “structural elements”.
-Amino acid distributions at individual position should not be
taken as independent of one another.
-Investigation of correlations between sequence positions in
protein family leads to decomposition of the protein into groups
of coevolving amino acids – “sectors”.
Hypothesis: the sectors are features of proteins structures and reflect
the evolutionary histories of their conserved biological properties.
S1A Family
Serine protease
Clan
Family
Sub-family
SA
SB
…
S1
S2
…
S1A
…
Catalytic triad – active site
rat trypsin
(3TGI)
Member
trypsin
chymotrypsin
tryptase
kallikrein
granzyme
…
Broad distribution and functions
Prokaryotes
Digestion
Invertebrates
Blood clotting
Vertebrates
Inflammation
…
Binding site - specificity
Method Outline

Identification of sectors


Statistical Coupling Analysis
Statistical Independence

Correlated entropy

Physical connectivity

Distinct biochemical properties



Alanine mutagenesis
Catalytic power & thermal stability assays
Independent divergence

Sequence similarity analysis
From Sequence to Sectors
Multiple sequence alignment of 1470 members of the S1A family
(single domain)
NCBI nonredundant database through iterative PSI-BLAST
Alignment: Cn3D, ClustalX
Standard manual adjustment methods
Position Conservation
Di(a): Divergence (or relative entropy)
fi(a): Observed frequency of amino acid a at position i
q(a): Background frequency of a in all proteins
Statistical Coupling Analysis (SCA)
SCA matrix (conservation-weighted covariance matrix)
Cijab: frequency-based correlation between position i and j
~Cijab is a measure of the significance of observed correlations as
judged by the conservation of the amino acids under consideration
After binary approximation:
Binary approximation
Di(ai): the conservation of
ai, which is the most
prevalent amino acid at
that position
Spectral cleaning to separate functional correlation
from statistical and historical noise
Principal Component Analysis
Spectral decomposition of ~Cij matrix
to partially sort out the different
contributions to the correlations
223 eigenvalues
Lowest 218 – Statistical noise
Randomized alignments retaining the
same size and amino acid propensities
at sites show eigenvalues of similar
magnitude
First mode makes the dominant
contribution to ~Cij – historical noise
The first eiganvelue is well approximated
by a first order approximation, proves that
the first eigenvector should just report the
net contribution of each position to the
total correlation
Sector Identification using modes 2 to 5
Overview of Sectors
Statistical Independence
Compute correlation entropy to
quantitatively measure the independence
of sectors
Minimum discriminatory information method
i.e.
S is small set of position,
specifically, the top five positions
contributing to each sector
Structure Connectivity
Known primary/secondary/subdomain-architecture subdivision
No sector
Distinction in degree of solvent exposure
Difference in proximity to the active site (not for green sector)
Red: focus on S1 pocket
catalytic specificity
Blue: more distributed property
Green: focus around catalytic triad
catalytic activity
Without information about tertiary structure and only ~10% of total sequence
positions contributes strongly to each sector, each sector reveals obvious intrasector physical connectivity and only a few inter-sector contacts.
Biochemical Independence
Mutations of red and blue sectors
showed very different effects focused
either on catalytic power or thermal
stability
Additive effects from combination of
mutations between two groups
(magenta: observed | white: predicted)
Independent Sequence Divergence
Sequence similarity analysis of each sector classifies members in the family
effectively only by the related property, while the analysis on all positions failed to
do the classification (442 members with functional annotation)
Evidence of “Sector” theory in
Other Protein Families
PDZ
PAS
SH2
SH3
Different regulatory
mechnisms
Discussion
Novel Structural Organization
Implication for Physical Properties of Proteins
Alternative View to Calculate Residue Covariance
Technical Challenges
Protein Modulization Adaptive Advantage