Biology Reference
In-Depth Information
identity level. Within each cluster, one representative structure
is selected. X-ray structures are preferred over NMR or
CryoEM structures. Among the X-ray structures, the one
with the lowest resolution is selected as the representative.
The representative structures that are classified in the SCOP
database are further split into structural domains according to
SCOP domain definitions.
8. The PROMALS3D Web server also offers an option to use
PROMALS [ 19 ] to align within each pre-aligned group instead
of MAFFT. This option is currently not available in the
downloadable package. PROMALS uses sequence homologs
and predicted secondary structures, and thus often produces
better alignment results. This is helpful to achieve better overall
alignment quality when pre-aligned groups themselves contain
divergent sequences. However, due to database searching
PROMALS is more time-consuming compared to MAFFT.
9. Three options of structural alignments and their combinations
are offered: DaliLite [ 35 ], FAST [ 36 ], and TM-align [ 37 ].
DaliLite gives slightly better results than FAST and TM-align
[ 23 ]. Using combinations of them also provides slight
improvement of alignment accuracy [ 23 ]. The default option
of the PROMALS3D Web server is the combination of FAST
and TMalign. DaliLite is computationally intensive when the
structures are large (e.g., with more than 500 residues).
10. In addition to input sequences and structures, PROMALS3D
also allows
input of alignment constraints
(user-defined
constraints).
11. While PROMALS3D compares favorably to a number of other
methods on an average basis [ 23 ], it does not mean it can
outperform any method for any individual alignment case. For
regions with uncertainty, inspection of results produced by other
methods could be helpful tomanually improve alignment quality.
12. The advantage of PROMALS3D is the incorporation of infor-
mation from homologous sequences and structures. However,
mistakes may be introduced in the process. For example, PSI-
BLAST may find nonhomologous sequences (profile corrupt),
and the PSI-BLAST alignment between the query and its hits
may contain errors that could lead to inferior profile or wrong
profile-profile alignment. The PSI-BLAST results of sequence
and structure database searches are kept and can be accessed
from the PROMALS3D Web server.
13. Alignment mistakes could also be caused by wrong secondary
structure predictions. While PSIPRED secondary structure
prediction accuracy is on average about 70-80 %, it is more
difficult to obtain accurate predictions for beta-strands and in
cases where few homologous sequences exist.
Search WWH ::




Custom Search