Biology Reference
In-Depth Information
d
a
Restrict the area of the DP matrix
Convert an amino acid sequence to a 2D wave
sequence 2
site 4
12
3
4
5
site 3
W
Y
site 1
site 2
W
R
K
F
Y
W
1
2
3
K
S
11
W
I
L
V
F
R
F
Y
Y
M
K
H
I
L
K
S
23
F
R
R
F
K
M
K
Q
E
I
L
V
H
LAFADKTNVK
I
L
L
S
32
M
T
QE
N
H
M
V
H
C
D
T
N
QE
T
V
QE
V
C
N
D
D
A
S
T
T
P
C
N
D
C
N
D
A
A
S
G
S
44
S
55
P
4
5
A
A
S
A
P
S
G
Polarity
P
G
G
Polarity
c
c
(
k
)
Correlation coefficient
b
Convert a profile to a 2D wave
sequence 2
k
24
0
site 4
site 3
W
site 1
Y
site 2
W
F
R
W
Y
W
K
I
L
R
F
Y
Y
M
K
H
I
L
R
R
F
F
WYDAERAAI
V--ADRAGV
K
Q
E
M
V
H
K
I
L
I
L
T
M
H
M
QE
V
H
C
N
D
T
QE
Q E
V
V
C
N
D
A
S
T
T
P
C
N
D
C
N
D
A
P
S
G
A
P
S
A
S
G
Polarity
P
G
G
Polarity
Fig. 2 Calculation procedure of the iterative refinement options (reprinted from [
57
])
In the group-to-group alignment calculation, each group can have
gaps introduced in previous progressive steps. These gap positions
are approximately considered [
1
].
In the FFT-NS-2 and FFT-NS-1 options, a group-to-group
alignment algorithm based on the FFT approximation [
1
] is used
(
see
Fig.
2
). It enables fast calculation for long sequences. However,
for handling many short sequences, sometimes the use of FFT is
not preferable, because it requires additional memory space and
conversion steps from sequence to wave. For such cases, standard
dynamic programming (DP) can be selected. These options are
called NW-NS-1 and NW-NS-2 for one-cycle and two-cycle ver-
sions, respectively.
It is possible to use pairwise alignments by DP, instead of 6mer
comparison, for calculating a guide tree.
L-INS-1 and G-INS-1 use local alignment and global alignment,
respectively, for all the pairs to compute a distance matrix. These
methods are slower but more accurate than FFT-NS-2 and NW-
NS-2 in most benchmarks.
