Biology Reference
In-Depth Information
Fig. 3 Schematic overview of the secondary structure-guided alignment strategy in PRALINE (Pirovano,
Simossis, and Heringa, unpublished). For details, see text
It is well known that the secondary structure elements of proteins
are much more conserved than their amino acid sequence during
evolution [
16
,
19
-
21
]. Therefore, secondary structure information
can be used to guide the alignment process, particularly in the case
of distantly related proteins [
7
,
8
,
18
,
22
-
27
].
The secondary structure-guided alignment strategy in PRA-
LINE works by combining secondary structure prediction with a
secondary structure-based scoring scheme (Fig.
3
). When using
predicted secondary structure, however, the gain in information
might be overshadowed by prediction error. Fortunately, during
earlier tests with the PRALINE secondary structure-guided strat-
egy, it turned out that the inclusion of secondary structure infor-
mation improves alignment whenever a prediction accuracy of 65 %
or more is achieved (Simossis and Heringa, unpublished), and this
is easily attained by modern prediction methods.
PRALINE starts the strategy by predicting the secondary struc-
ture elements of each sequence using a secondary structure predic-
tion tool. PRALINE provides the user with the choice of four
different secondary structure predictors: PSIPRED [
28
], SSPRO
4.0 [
29
], PORTER [
30
], and YASPIN [
31
]. Each of these predic-
tors has its own strengths and weaknesses, the choice of which is
therefore left to the user's discretion. The secondary structure
prediction methods perform a PSI-BLAST search for each input
sequence and then perform the secondary structure prediction
using the position-dependent scoring matrix (PSSM) produced by
PSI-BLAST, thereby making use of the amino acid conservation as
observed in the putative homologous sequences. If an input
2.4 Secondary
Structure-Guided
Alignment
