Table 1: The following table contains some localization examples where
AtSubP predicts correctly; whereas
TargetP, the widely used prediction system (e.g. by
TAIR,
MIPS,
PLASdb etc. for Arabidopsis annotation), gives wrong or no prediction. Please note: the following list is not exhaustive as we include only few examples from the
experimentally-proved sequences available at the Arabidopsis Subcellular Database (
SUBA). Moreover, to increase the confidence level, we further filter this list containing sequences NOT used in the training of AtSubP system; i.e. the GFP data for TAIR IDs numbered 1 - 17 are taken from SUBA (also present in our
independent test set-II). Similarly, we take some examples (TAIR IDs 18 - 23) from Dr. Niko Geldner's wave list published recently (
Geldner et al., 2009).
Furthermore,
AtSubP has been used by some of the researchers at our Foundation in their own projects and performed
green fluorescent protein (GFP) fusions on such previously '
unknown' proteins (TAIR IDs 24 - 26). Interestingly, these proteins were also wrongly predicted by
TargetP, and accordingly, we have included them in the following table.