Table 2.

Glimmer3 prediction accuracy when trained on confirmed genes

Genome			Glimmer3 Predictions					versus Glimmer2.13
Organism	GC%	# Genes	3′ Matches		5′ & 3′ Matches		Extra	3′ Match	5′ & 3′	Extra
A.fulgidus	49	1165	1162	99.7%	841	72.2%	1308	−2	−67	−59
B.anthracis	35	3132	3119	99.6%	2717	86.7%	2345	+6	+726	−77
B.subtilis	44	1576	1559	98.9%	1379	87.5%	2886	+11	+413	−539
C.tepidum	57	1292	1284	99.4%	867	67.1%	778	+2	−33	−190
C.perfringens	29	1504	1501	99.8%	1360	90.4%	1177	−1	+244	−28
E.coli	51	3603	3525	97.8%	3014	83.7%	942	+16	+693	−632
G.sulfurreducens	61	2351	2320	98.7%	1883	80.1%	1107	+15	+541	−380
H.pylori	39	915	908	99.2%	785	85.8%	774	+1	+46	−94
P.fluorescens	63	4535	4484	98.9%	3412	75.2%	1896	+14	+731	−704
R.solanacearum	67	2512	2468	98.2%	1922	76.5%	1091	+72	+646	−326
S.epidermidis	32	1650	1646	99.8%	1496	90.7%	767	+3	+338	−66
T.pallidum	53	575	569	99.0%	397	69.0%	568	+3	+55	−296
U.parvum	26	327	325	99.4%	292	89.3%	297	0	+19	−17
Averages:				99.1%		81.1%		+11	+335	−262

For each genome, ‘#Genes’ counts genes in the NCBI annotation that are at least 90bp long, do not have frame shifts or internal stop codons, and whose function description does not contain the string ‘hypothetical’. Both Glimmer3 and Glimmer2 were run with the same options and training/test sets in an 8-way cross-validation experiment on this set of genes. A prediction is a 3′ match iff it has the same reading frame and stop codon as a gene. 5′ & 3′ matches are predictions with the same start and stop codon as the annotation. Extra predictions are those that are not matches. The ‘versus Glimmer2’ column is the Glimmer3 value minus the corresponding Glimmer2 value.