Biomedical Engineering Reference
In-Depth Information
9. The sequence variability of all parts of the
b
-barrel during evolution is high
when compared with soluble proteins.
10. The external loops show exceptionally high sequence variability and they are
usually mobile.
3.4 Discrimination of
-Barrel Membrane Proteins
b
Several methods have been proposed for discriminating TMBs from other folding
types of globular and membrane proteins. These methods are broadly classified into
two aspects: statistical analyses and machine-learning techniques.
3.4.1 Statistical Methods
The statistical methods are straightforward and these methods could explain their
merits and limitations based on physical principles. In statistical methods, amino
acid composition, residue pair preference and motifs have been mainly used for
discrimination. In addition, hydrophobicity profiles, composition of residues in the
membrane environment, and sequence alignments have been used for discriminat-
ing TMBs.
Gnanasekaran et al. [
40
] developed a structure-based sequence alignment
method for discriminating
MBs and reported an accuracy of 80% for a limited
number of proteins. Wimley [
41
] analyzed the architecture of 15 TMBs and
proposed a method based on hydrophobicity for identifying
Τ
MBs in genomic
sequences. It has been reported that this method correctly identified 75% of the
TMBs. Liu et al. [
42
] analyzed the amino acid composition in the membrane
spanning regions of 12
Τ
MBs and applied the information for discrimination,
which showed 85% accuracy when tested with 241 TMBs. We have proposed
methods using the compositions of amino acids, residue pairs, and motifs
[
43
-
45
]. In these methods, the compositions have been computed for both positive
and negative datasets (TMS proteins and other globular/TMH proteins, respec-
tively) and treated as standard compositions. For a new protein, we computed the
composition and compared with standard compositions (positive and negative
datasets). The test protein is assigned to TMB whether deviation is smallest from
the standard composition for TMB and vice versa. The discrimination based on
residue pair preference is explained below:
(1) Calculate the dipeptide composition for both globular (Dipep
glob
) and TMBs
(Dipep
TMB
) and the difference between them (
Τ
s
TMB-glob
); (2) for a new protein, X,
calculate the dipeptide composition and give weights to the dipeptide composition
with
s
TMB-glob
; (3) calculate the sum of weighted dipeptide composition, and (4)
the protein X is predicted to be an TMB if the total weighted dipeptide composition
is positive and globular protein otherwise [
44
]. For a set of 377 TMBs and 674
globular proteins, this method correctly identified 357 TMBs (95%) and excluded
534 globular proteins (79%). The twofold cross validation test showed an average
