AVID: An integrative framework for discovering functional relationships among proteins

BMC Bioinformatics

Table 1 Correlation between experimental or sequence-based measures of protein relatedness and GO annotation similarity.

Data^a	MF^b	BP^b	CC^b
UCSF localization (+)	0.031	0.029	0.058
UCSF localization (-)	0.007	0.007	0.004
yeast 2-hybrid (+)	0.120	0.167	0.254
yeast 2-hybrid (-)	0.010	0.010	0.010
MIPS complex (+)	0.122	0.116	0.263
MIPS complex (-)	0.007	0.008	0.002
sequence similarity (+)	0.187	0.344	0.078
sequence similarity (-)	0.009	0.007	0.010
Microarray^c
-1 < R ≤ -0.9	0.008	0.003	0.006
0.9 < R ≤ 1.0	0.086	0.118	0.125

^aExcept for the microarray data, each data source is divided into two sets: one contains protein pairs observed to share the feature described by the data (+), the other contains protein pairs that lack the feature or are not reported (-). ^bThe normalized conditional probabilities P_ij>^AVID1 are defined in the Methods; they are the probability of sharing a common GO term given observation of a particular data feature. A perfect correlation would give a value of 1.0. ^cR is the Pearson correlation coefficient for pairs of mRNA expression profiles, which were binned into 19 intervals. Only two intervals are shown here. See additional file 1 for further details.

ISSN: 1471-2105