Programmes de détection des gènes Microbiens

Les ORF (Open Reading Frames, cadres de lecture ouverts) supérieurs à 150 nt. sont pratiquement tous vrais
Problème pour les ORF courts, inf. à 150.

GENEMARK. Utilise les modeles markoviens dans sa version de base (recherche des regions codantes seulement). Puis l'idee a ete integree dans un modèle markovien cache (HMM) plus complexe tenant compte des informations du promoteur et du RBS.
ECOPARSE: HMM pour la détection des gènes de E. coli.

Document:

Results of GeneMark.hmm predictions for 10 complete bacterial genomes*

Genome	Genes annotated	Genes predicted	Annotated genes predicted by GeneMark.hmm & GeneMark (%)	Correct 5' end prediction of annotated genes (%)	Potential new genes (%)
A.fulgidus	2407	2530	98.0	73.1	15.1
B.subtilis	4101	4384	97.2	77.5	9.8
E.coli	4288	4440	97.3	75.4	8.2
H.influenzae	1718	1840	96.2	86.7	10.2
H.pylori	1566	1612	95.6	79.7	8.7
M.genitalium	467	509	98.3	78.4	17.3
M.jannaschii	1680	1841	99.2	72.7	12.9
M.pneumoniae	678	734	95.9	70.1	13.6
M.thermoauto	1869	1944	97.5	70.9	8.6
Synechocystis	3169	3360	98.5	89.6	9.4
Average	21943	23194	97.3	78.1	10.4

The second and third columns show the number of genes annotated in GenBank and the number of genes predicted, respectively.
The "Annotated genes predicted" column presents the percentage of annotated genes which were predicted by GeneMark and GeneMark.hmm
The "Correct 5' end prediction of annotated genes" column shows the percentage of genes whose starts were predicted exactly.
"Potential new genes" is the fraction of predicted genes for which no annotated analog was found. All measures are expressed in percent.
* Reference: A. Lukashin and M. Borodovsky, GeneMark.hmm: new solutions for gene finding, NAR, 1998, Vol. 26, No.4, pp 1107-1115.