Table 1

Results from computational approaches predicting nuclear proteome membership

Method
NUCPROT proteins classified as 'nuclear'
Accuracy
RIKEN proteome proteins classified as 'nuclear'

CELLO
2,125
78%
9,122 (47%)
pTARGET
1,706
63%
5,953 (30%)
Proteome Analyst
1,803
66%
4,084 (21%)
WoLF PSORT
1,909
70%
7,172 (37%)
MultiLoc
1,561
57%
5,137 (26%)




Yeast homology (E < 10-4)
218
8.0%
2,031 (8.0%)
Yeast homology (E < 10-30)
47
1.7%
691 (3.5%)
Mouse homology
430
16%
766 (3.9%)




Nucleo
857
32%
987 (5.0%)

For the subcellular localization prediction programs, proteins were considered to be incorrectly classified as 'not nuclear' if the method's top-ranked localization call was not 'nucleus' but the protein was in our high-quality dataset; proteins were considered to be correctly classified as 'nuclear' if the method's top-ranked localization call was 'nuclear' and the protein was in our high-quality dataset.

Fink et al. Genome Biology 2008 9:R15   doi:10.1186/gb-2008-9-1-r15