<s>
GeneMark	B-Application
is	O
a	O
generic	O
name	O
for	O
a	O
family	O
of	O
ab	O
initio	O
gene	O
prediction	O
programs	O
developed	O
at	O
the	O
Georgia	O
Institute	O
of	O
Technology	O
in	O
Atlanta	O
.	O
</s>
<s>
Developed	O
in	O
1993	O
,	O
original	O
GeneMark	B-Application
was	O
used	O
in	O
1995	O
as	O
a	O
primary	O
gene	O
prediction	O
tool	O
for	O
annotation	O
of	O
the	O
first	O
completely	O
sequenced	O
bacterial	O
genome	O
of	O
Haemophilus	O
influenzae	O
,	O
and	O
in	O
1996	O
for	O
the	O
first	O
archaeal	O
genome	O
of	O
Methanococcus	O
jannaschii	O
.	O
</s>
<s>
Original	O
GeneMark	B-Application
(	O
developed	O
before	O
the	O
HMM	O
era	O
in	O
Bioinformatics	O
)	O
is	O
an	O
HMM-like	O
algorithm	O
;	O
it	O
can	O
be	O
viewed	O
as	O
approximation	O
to	O
known	O
in	O
the	O
HMM	O
theory	O
posterior	O
decoding	O
algorithm	O
for	O
appropriately	O
defined	O
HMM	O
.	O
</s>
<s>
The	O
GeneMark.hmm	O
algorithm	O
(	O
1998	O
)	O
was	O
designed	O
to	O
improve	O
gene	O
prediction	O
accuracy	O
in	O
finding	O
short	O
genes	O
and	O
gene	O
starts	O
.	O
</s>
<s>
The	O
idea	O
was	O
to	O
integrate	O
the	O
Markov	O
chain	O
models	O
used	O
in	O
GeneMark	B-Application
into	O
a	O
hidden	O
Markov	O
model	O
framework	O
,	O
with	O
transition	O
between	O
coding	O
and	O
non-coding	O
regions	O
formally	O
interpreted	O
as	O
transitions	O
between	O
hidden	O
states	O
.	O
</s>
<s>
Next	O
step	O
was	O
done	O
with	O
development	O
of	O
the	O
self-training	O
gene	O
prediction	O
tool	O
GeneMarkS	B-Application
(	O
2001	O
)	O
.	O
</s>
<s>
GeneMarkS	B-Application
has	O
been	O
in	O
active	O
use	O
by	O
genomics	O
community	O
for	O
gene	O
identification	O
in	O
new	O
prokaryotic	O
genomic	O
sequences	O
.	O
</s>
<s>
GeneMarkS+	O
,	O
extension	O
of	O
GeneMarkS	B-Application
integrating	O
information	O
on	O
homologous	O
proteins	O
into	O
gene	O
prediction	O
is	O
used	O
in	O
the	O
NCBI	O
pipeline	O
for	O
prokaryotic	O
genomes	O
annotation	O
;	O
the	O
pipeline	O
can	O
annotate	O
up	O
to	O
2000	O
genomes	O
daily	O
(	O
)	O
.	O
</s>
<s>
Accurate	O
identification	O
of	O
species	O
specific	O
parameters	O
of	O
the	O
GeneMark	B-Application
and	O
GeneMark.hmm	O
algorithms	O
was	O
the	O
key	O
condition	O
for	O
making	O
accurate	O
gene	O
predictions	O
.	O
</s>
<s>
The	O
HMM	O
architecture	O
of	O
eukaryotic	O
GeneMark.hmm	O
includes	O
hidden	O
states	O
for	O
initial	O
,	O
internal	O
,	O
and	O
terminal	O
exons	O
,	O
introns	O
,	O
intergenic	O
regions	O
and	O
single	O
exon	O
genes	O
located	O
in	O
both	O
DNA	O
strands	O
.	O
</s>
<s>
Initial	O
eukaryotic	O
GeneMark.hmm	O
needed	O
training	O
sets	O
for	O
estimation	O
of	O
the	O
algorithm	O
parameters	O
.	O
</s>
<s>
In	O
2005	O
first	O
version	O
of	O
self-training	O
algorithm	O
GeneMark-ES	O
was	O
developed	O
.	O
</s>
<s>
In	O
2008	O
the	O
GeneMark-ES	O
algorithm	O
was	O
extended	O
to	O
fungal	O
genomes	O
by	O
developing	O
a	O
special	O
intron	O
model	O
and	O
more	O
complex	O
strategy	O
of	O
self-training	O
.	O
</s>
<s>
Then	O
,	O
in	O
2014	O
,	O
GeneMark-ET	O
the	O
algorithm	O
that	O
augmented	O
self-training	O
by	O
information	O
from	O
mapped	O
to	O
genome	O
unassembled	O
RNA-Seq	O
reads	O
was	O
added	O
to	O
the	O
family	O
.	O
</s>
<s>
GeneMark-ES	O
:	O
gene	O
finding	O
algorithm	O
for	O
eukaryotic	O
genomes	O
that	O
performs	O
automatic	O
training	O
in	O
unsupervised	O
ab	O
initio	O
mode	O
.	O
</s>
<s>
GeneMark-ET	O
:	O
augments	O
GeneMark-ES	O
with	O
a	O
novel	O
method	O
that	O
integrates	O
RNA-Seq	O
read	O
alignments	O
into	O
the	O
self-training	O
procedure	O
.	O
</s>
<s>
GeneMark-EX	O
:	O
a	O
fully	O
automatic	O
integrated	O
tool	O
for	O
genome	O
annotation	O
that	O
shows	O
robust	O
performance	O
across	O
the	O
input	O
data	O
of	O
various	O
size	O
,	O
structure	O
and	O
quality	O
.	O
</s>
