We installed New gene-finding HMM based program FGENESH
for multiple gene prediction in human (mammalian) genomic DNA
It shows superior accuracy of gene prediction
(Salamov A.A, Solovyev V.V, 1999)
Very important feature!!!! If we take the subset of exons predicted by
Fgenes and Fgenesh (programs using very different approaches)
for long genomic sequences or sequences with multiple genes,
then the observed specificity is 95% and this set will include about
76% of all real exons. These predictions will provide positional
cloners, gene hunters and others with the best exon candidates observed
in finished and unfinished genomic sequences.
It is available at http://genomic.sanger.ac.uk/ of our
Computational Genomic Group WEB server
(http://genomic.sanger.ac.uk/gf/gf.html)
Accuracy of gene-finding programs for 1.4 MB genomic region on human chromosome
13q
=================================================================================
*** accuracy data provided by Tim Habburd and Richard Briskevich (the Sanger
Centre)
as a blind test for difficult genomic region
acc -is specificity (true predicted/all predicted) cov is sensitivity (true
predicted/true)
NE - number of predicted exons
Region includes 159 true exons
exact match overlap exons 5'-exons
3'-exons
NE N acc cov N acc cov N acc cov
N acc cov
fgenesh.masked 169 110 0.65 0.69 125 0.74 0.79 118 0.70 0.74
116 0.69 0.73
fgenesh 190 109 0.57 0.69 126 0.66 0.79 117 0.62 0.74
117 0.62 0.74
fgenes.masked 238 103 0.43 0.65 132 0.55 0.83 114 0.48 0.72
118 0.50 0.74
fgenes 281 104 0.37 0.65 136 0.48 0.86 116 0.41 0.73
120 0.43 0.75
genscan 292 105 0.36 0.66 129 0.44 0.81 116 0.40 0.73
115 0.39 0.72
fgeneh 381 68 0.18 0.43 101 0.27 0.64 79 0.21 0.50
87 0.23 0.55
mzef 623 95 0.15 0.60 122 0.20 0.77 106 0.17 0.67
107 0.17 0.67
***
--
Victor Solovyev
The Sanger Centre, Hinxton, Cambridge CB10 1SA, UK
Email: solovyev at sanger.ac.ukhttp://genomic.sanger.ac.uk
Phone: 44-1223-494799 FAX: 44-1223-494919
