<s>
In	O
bioinformatics	O
,	O
GENSCAN	B-Application
is	O
a	O
program	B-Application
to	O
identify	O
complete	O
gene	O
structures	O
in	O
genomic	O
DNA	O
.	O
</s>
<s>
It	O
is	O
a	O
GHMM-based	O
program	B-Application
that	O
can	O
be	O
used	O
to	O
predict	O
the	O
location	O
of	O
genes	O
and	O
their	O
exon-intron	O
boundaries	O
in	O
genomic	O
sequences	O
from	O
a	O
variety	O
of	O
organisms	O
.	O
</s>
<s>
The	O
GENSCAN	B-Application
Web	O
server	O
can	O
be	O
found	O
at	O
MIT	O
.	O
</s>
<s>
GENSCAN	B-Application
was	O
developed	O
by	O
Christopher	O
Burge	O
in	O
the	O
research	O
group	O
of	O
Samuel	O
Karlin	O
at	O
Stanford	O
University	O
.	O
</s>
<s>
This	O
resulted	O
in	O
the	O
development	O
of	O
a	O
program	B-Application
called	O
TWINSCAN	O
as	O
an	O
adaptation	O
of	O
GENSCAN	B-Application
with	O
higher	O
accuracy	O
.	O
</s>
<s>
As	O
of	O
2002	O
,	O
GENSCAN	B-Application
remained	O
a	O
popular	O
tool	O
in	O
bioinformatics	O
,	O
becoming	O
a	O
standard	O
feature	O
for	O
genomes	O
released	O
on	O
University	O
of	O
California	O
Santa	O
Cruz	O
and	O
Ensembl	O
Genome	O
browser	O
.	O
</s>
<s>
The	O
primary	O
goal	O
when	O
developing	O
a	O
genomic	O
sequence	O
model	O
for	O
GENSCAN	B-Application
was	O
to	O
identify	O
both	O
the	O
general	O
and	O
specific	O
properties	O
that	O
compose	O
the	O
individual	O
functional	O
units	O
of	O
eukaryotic	O
genes	O
(	O
e.g.	O
</s>
<s>
Due	O
to	O
the	O
usage	O
of	O
these	O
elements	O
,	O
GENSCAN	B-Application
works	O
without	O
needing	O
to	O
reference	O
similar	O
genes	O
in	O
protein	O
sequence	O
databases	O
.	O
</s>
<s>
Instead	O
,	O
predictions	O
produced	O
by	O
GENSCAN	B-Application
are	O
complementary	O
to	O
those	O
gathered	O
by	O
homology-based	O
gene	O
identification	O
methods	O
(	O
e.g.	O
</s>
<s>
querying	O
protein	O
databases	O
with	O
BLASTX	B-Application
)	O
.	O
</s>
<s>
Overall	O
,	O
the	O
structure	O
of	O
the	O
model	O
used	O
in	O
GENSCAN	B-Application
is	O
similar	O
to	O
the	O
General	O
Hidden	O
Markov	O
Model	O
.	O
</s>
<s>
GENSCAN	B-Application
's	O
implementation	O
differs	O
from	O
other	O
programs	O
in	O
multiple	O
ways	O
.	O
</s>
<s>
A	O
notable	O
difference	O
is	O
the	O
fact	O
that	O
GENSCAN	B-Application
utilizes	O
a	O
genomic	O
sequence	O
model	O
that	O
exclusively	O
focuses	O
double-stranded	O
DNA	O
where	O
genes	O
that	O
are	O
present	O
on	O
both	O
strands	O
are	O
simultaneously	O
analyzed	O
.	O
</s>
<s>
Also	O
,	O
GENSCAN	B-Application
is	O
capable	O
of	O
analyzing	O
genomes	O
in	O
situations	O
where	O
there	O
are	O
partial	O
genes	O
or	O
no	O
genes	O
,	O
rather	O
than	O
only	O
being	O
able	O
to	O
analyze	O
single	O
and	O
complete	O
gene	O
sequences	O
like	O
other	O
programs	O
at	O
its	O
time	O
.	O
</s>
<s>
These	O
two	O
factors	O
contribute	O
to	O
GENSCAN	B-Application
being	O
particularly	O
useful	O
in	O
analyzing	O
longer	O
human	O
genomes	O
.	O
</s>
<s>
In	O
addition	O
,	O
GENSCAN	B-Application
employs	O
the	O
concept	O
of	O
Maximal	O
Dependence	O
Decomposition	O
such	O
that	O
functional	O
signals	O
in	O
DNA	O
and	O
protein	O
sequences	O
can	O
be	O
modeled	O
,	O
creating	O
the	O
possibility	O
for	O
dependencies	O
between	O
signal	O
positions	O
to	O
be	O
considered	O
by	O
the	O
program	B-Application
.	O
</s>
<s>
This	O
is	O
implemented	O
in	O
GENSCAN	B-Application
such	O
that	O
a	O
model	O
is	O
generated	O
of	O
the	O
donor	O
splice	O
signal	O
,	O
capturing	O
dependences	O
that	O
are	O
associated	O
with	O
the	O
recognition	O
mechanisms	O
for	O
donor	O
splice	O
sites	O
in	O
pre-mRNA	O
sequences	O
.	O
</s>
<s>
GENSCAN	B-Application
has	O
the	O
capability	O
of	O
calculating	O
the	O
accuracy	O
of	O
each	O
of	O
its	O
predictions	O
by	O
using	O
the	O
forward-backward	B-Algorithm
algorithm	I-Algorithm
.	O
</s>
<s>
Predicting	O
the	O
structure	O
and	O
overall	O
composition	O
of	O
human	O
genes	O
in	O
regard	O
to	O
exon	O
and	O
gene	O
locations	O
in	O
longer	O
sequences	O
is	O
an	O
additionally	O
useful	O
component	O
of	O
GENSCAN	B-Application
.	O
</s>
<s>
Lastly	O
,	O
this	O
also	O
allows	O
GENSCAN	B-Application
to	O
capture	O
dependencies	O
between	O
signal	O
positions	O
with	O
new	O
models	O
of	O
donor	O
and	O
acceptor	O
splice	O
sites	O
.	O
</s>
<s>
The	O
run	O
time	O
for	O
GENSCAN	B-Application
scales	O
almost	O
linearly	O
when	O
provided	O
realistically	O
sized	O
sequences	O
(	O
several	O
kilobits	O
minimum	O
)	O
,	O
but	O
has	O
a	O
worst	O
case	O
of	O
being	O
quadratic	O
.	O
</s>
<s>
GENSCAN	B-Application
,	O
like	O
other	O
genome	O
prediction	O
programs	O
,	O
does	O
n't	O
produce	O
results	O
that	O
totally	O
match	O
those	O
of	O
other	O
programs	O
.	O
</s>
<s>
Therefore	O
,	O
GENSCAN	B-Application
has	O
been	O
utilized	O
in	O
the	O
practice	O
of	O
combining	O
two	O
gene	O
prediction	O
programs	O
 '	O
results	O
such	O
that	O
if	O
one	O
program	B-Application
in	O
the	O
combination	O
is	O
confident	O
in	O
a	O
sequence	O
prediction	O
,	O
that	O
sequence	O
is	O
used	O
.	O
</s>
<s>
On	O
the	O
other	O
hand	O
,	O
if	O
neither	O
program	B-Application
is	O
confident	O
in	O
their	O
predictions	O
,	O
the	O
sequence	O
predicted	O
is	O
only	O
used	O
if	O
both	O
programs	O
agree	O
on	O
it	O
.	O
</s>
<s>
Tests	O
were	O
conducted	O
to	O
evaluate	O
the	O
accuracy	O
of	O
GENSCAN	B-Application
with	O
short	O
data	O
sets	O
.	O
</s>
<s>
GENSCAN	B-Application
is	O
shown	O
in	O
the	O
table	O
to	O
be	O
generally	O
more	O
accurate	O
than	O
its	O
competitors	O
at	O
predicting	O
sequences	O
with	O
both	O
nucleotides	O
and	O
exons	O
.	O
</s>
<s>
Furthermore	O
,	O
the	O
table	O
shown	O
below	O
specifically	O
describes	O
the	O
accuracy	O
of	O
GENSCAN	B-Application
in	O
regard	O
to	O
genomic	O
sequences	O
organized	O
by	O
ranges	O
of	O
C	O
+	O
G	O
and	O
types	O
of	O
organisms	O
.	O
</s>
<s>
We	O
can	O
see	O
in	O
the	O
data	O
provided	O
that	O
GENSCAN	B-Application
's	O
accuracy	O
variation	O
was	O
rather	O
insensitive	O
to	O
C	O
+	O
G	O
content	O
and	O
organism	O
type	O
.	O
</s>
<s>
This	O
further	O
demonstrates	O
GENSCAN	B-Application
's	O
independence	O
of	O
factors	O
that	O
would	O
have	O
impacted	O
the	O
results	O
of	O
comparable	O
genome	O
prediction	O
programs	O
.	O
</s>
<s>
+GENSCAN	O
Accuracy	O
for	O
Sequences	O
Organized	O
by	O
C+G	O
Content	O
and	O
OrganismSubsetSequencesNucleotide	O
SensitivityNucleotide	O
SpecificityNucleotide	O
Approximate	O
CorrelationNucleotide	O
Correlation	O
CoefficientExon	O
SensitivityExon	O
SpecificityExon	O
AverageMissed	O
ExonsWrong	O
ExonsC	O
+	O
G	O
<	O
40860.900.950.900.930.780.870.840.140.05C	O
+	O
G	O
40-502200.940.920.910.910.800.820.820.080.05C	O
+	O
G	O
50-602080.930.930.900.920.750.770.770.080.05C	O
+	O
G	O
>60560.970.890.900.900.760.770.760.070.08Primates2370.960.940.930.940.810.820.820.070.05Rodents1910.900.930.890.910.750.800.780.110.05Non-mamm	O
.	O
</s>
<s>
A	O
separate	O
test	O
was	O
conducted	O
on	O
GENSCAN	B-Application
's	O
accuracy	O
using	O
two	O
GeneParser	O
data	O
sets	O
that	O
are	O
stripped	O
of	O
all	O
genes	O
that	O
are	O
more	O
than	O
25%	O
of	O
a	O
match	O
regarding	O
amino	O
acids	O
with	O
those	O
in	O
previous	O
GeneParser	O
test	O
sets	O
.	O
</s>
<s>
We	O
can	O
see	O
that	O
there	O
is	O
little	O
variation	O
between	O
the	O
accuracy	O
of	O
GENSCAN	B-Application
under	O
the	O
aforementioned	O
Burset/Guigó	O
data	O
set	O
and	O
the	O
GeneParser	O
data	O
sets	O
.	O
</s>
<s>
In	O
1997	O
,	O
GENSCAN	B-Application
was	O
found	O
to	O
have	O
a	O
higher	O
accuracy	O
than	O
previous	O
gene	O
prediction	O
programs	O
.	O
</s>
<s>
However	O
,	O
work	O
still	O
needed	O
to	O
be	O
done	O
due	O
to	O
how	O
GENSCAN	B-Application
was	O
shown	O
to	O
only	O
predict	O
10-15	O
%	O
of	O
genes	O
accurately	O
on	O
realistic	O
data	O
sets	O
.	O
</s>
<s>
Because	O
of	O
inaccuracies	O
like	O
this	O
,	O
any	O
predictions	O
given	O
by	O
GENSCAN	B-Application
and	O
other	O
programs	O
must	O
be	O
verified	O
by	O
comparing	O
them	O
to	O
a	O
Complementary	O
DNA	O
sequence	O
,	O
a	O
Expressed	O
sequence	O
tag	O
(	O
EST	O
)	O
sequence	O
,	O
or	O
a	O
known	O
protein	O
sequence	O
.	O
</s>
