<s>
(	O
Context-Specific	O
BLAST	B-Application
)	O
is	O
a	O
tool	O
that	O
searches	O
a	O
protein	O
sequence	O
that	O
extends	O
BLAST	B-Application
(	O
Basic	O
Local	O
Alignment	O
Search	O
Tool	O
)	O
,	O
using	O
context-specific	O
mutation	O
probabilities	O
.	O
</s>
<s>
More	O
specifically	O
,	O
CS-BLAST	B-Application
derives	O
context-specific	O
amino-acid	O
similarities	O
on	O
each	O
query	O
sequence	O
from	O
short	O
windows	O
on	O
the	O
query	O
sequences	O
 [ 4 ] 	O
.	O
</s>
<s>
Using	O
CS-BLAST	B-Application
doubles	O
sensitivity	O
and	O
significantly	O
improves	O
alignment	O
quality	O
without	O
a	O
loss	O
of	O
speed	O
in	O
comparison	O
to	O
BLAST	B-Application
.	O
</s>
<s>
CSI-BLAST	B-Application
(	O
Context-Specific	O
Iterated	O
BLAST	B-Application
)	O
is	O
the	O
context-specific	O
analog	O
of	O
PSI-BLAST	O
(	O
Position-Specific	O
Iterated	O
BLAST	B-Application
)	O
,	O
which	O
computes	O
the	O
mutation	O
profile	O
with	O
substitution	O
probabilities	O
and	O
mixes	O
it	O
with	O
the	O
query	O
profile	O
 [ 2 ] 	O
.	O
</s>
<s>
CSI-BLAST	B-Application
(	O
Context-Specific	O
Iterated	O
BLAST	B-Application
)	O
is	O
the	O
context	O
specific	O
analog	O
of	O
PSI-BLAST	O
(	O
Position-Specific	O
Iterated	O
BLAST	B-Application
)	O
.	O
</s>
<s>
Standard	O
methods	O
of	O
sequence	O
comparisons	O
use	O
a	O
substitution	B-Algorithm
matrix	I-Algorithm
to	O
accomplish	O
this	O
 [ 4 ] 	O
.	O
</s>
<s>
CS-BLAST	B-Application
improves	O
further	O
upon	O
this	O
concept	O
.	O
</s>
<s>
Substitution	B-Algorithm
matrix	I-Algorithm
scores	O
are	O
equivalent	O
to	O
profile	O
scores	O
if	O
the	O
sequence	O
profile	O
(	O
colored	O
histogram	O
)	O
is	O
generated	O
from	O
the	O
query	O
sequence	O
by	O
adding	O
artificial	O
mutations	O
with	O
the	O
substitution	B-Algorithm
matrix	I-Algorithm
pseudocount	O
scheme	O
.	O
</s>
<s>
CS-BLAST	B-Application
greatly	O
improves	O
alignment	O
quality	O
over	O
the	O
entire	O
range	O
of	O
sequence	O
identities	O
and	O
especially	O
for	O
difficult	O
alignments	O
in	O
comparison	O
to	O
regular	O
BLAST	B-Application
and	O
PSI-BLAST	O
.	O
</s>
<s>
PSI-BLAST	O
(	O
Position-Specific	O
Iterated	O
BLAST	B-Application
)	O
runs	O
at	O
about	O
the	O
same	O
speed	O
per	O
iteration	O
as	O
regular	O
BLAST	B-Application
,	O
but	O
is	O
able	O
to	O
detect	O
weaker	O
sequence	O
similarities	O
that	O
are	O
still	O
biologically	O
relevant	O
 [ 3 ] 	O
.	O
</s>
<s>
The	O
benchmark	O
compares	O
CS-BLAST	B-Application
to	O
BLAST	B-Application
using	O
true	O
positives	O
from	O
the	O
same	O
superfamily	O
versus	O
false	O
positive	O
of	O
pairs	O
from	O
different	O
folds	O
 [ 4 ] 	O
.	O
</s>
<s>
The	O
other	O
graph	O
uses	O
detects	O
true	O
positives	O
(	O
with	O
a	O
different	O
scale	O
than	O
the	O
previous	O
graph	O
)	O
and	O
false	O
positives	O
of	O
PSI-BLAST	O
and	O
CSI-BLAST	B-Application
and	O
compares	O
the	O
two	O
for	O
one	O
to	O
five	O
iterations	O
 [ 4 ] 	O
.	O
</s>
<s>
CS-BLAST	B-Application
offers	O
improved	O
sensitivity	O
and	O
alignment	O
quality	O
in	O
sequence	O
comparison	O
.	O
</s>
<s>
Sequence	O
searches	O
with	O
CS-BLAST	B-Application
are	O
more	O
than	O
twice	O
as	O
sensitive	O
as	O
BLAST	B-Application
 [ 4 ] 	O
.	O
</s>
<s>
CS-BLAST	B-Application
detects	O
139%	O
more	O
homologous	O
proteins	O
at	O
a	O
cumulative	O
error	O
rate	O
of	O
20%	O
 [ 2 ] 	O
.	O
</s>
<s>
At	O
a	O
10%	O
error	O
rate	O
,	O
138%	O
more	O
homologs	O
are	O
detected	O
,	O
and	O
for	O
the	O
easiest	O
cases	O
at	O
a	O
1%	O
error	O
rate	O
,	O
CS-BLAST	B-Application
was	O
still	O
96%	O
more	O
effective	O
than	O
BLAST	B-Application
 [ 2 ] 	O
.	O
</s>
<s>
Additionally	O
,	O
CS-BLAST	B-Application
in	O
2	O
iterations	O
is	O
more	O
sensitive	O
than	O
5	O
iterations	O
of	O
PSI-BLAST	O
.	O
</s>
<s>
The	O
CS-BLAST	B-Application
method	O
derives	O
similarities	O
between	O
sequence	O
context-specific	O
amino	O
acids	O
for	O
13	O
residue	O
windows	O
centered	O
on	O
each	O
residue	O
.	O
</s>
<s>
CS-BLAST	B-Application
works	O
by	O
generating	O
a	O
sequence	O
profile	O
for	O
a	O
query	O
sequence	O
by	O
using	O
context-specific	O
mutations	O
and	O
then	O
jumpstarting	O
a	O
profile-to-sequence	O
search	O
method	O
.	O
</s>
<s>
CS-BLAST	B-Application
starts	O
by	O
predicting	O
the	O
expected	O
mutation	O
probabilities	O
for	O
each	O
position	O
.	O
</s>
<s>
The	O
MPI	O
Bioinformatics	O
toolkit	O
in	O
an	O
interactive	O
website	O
and	O
service	O
that	O
allows	O
anyone	O
to	O
do	O
comprehensive	O
and	O
collaborative	O
protein	O
analysis	O
with	O
a	O
variety	O
of	O
different	O
tools	O
including	O
CS-BLAST	B-Application
as	O
well	O
as	O
PSI-BLAST	O
 [ 1 ] 	O
.	O
</s>
