Effects of protein interaction data integration, representation and reliability on the use of network properties for drug target prediction

Table 6 Average change in degree for drug targets and non-drug targets after removing lower-confidence interactions

Data sets	Avg degree change (drug targets)	Avg degree change (non-drug targets)	Wilcoxon p-value
Full to non- predicted	−5.2	−2.7	3.7e-34
Full to B	−6.2	−4.9	9.3e-13
Full to LTP	−10.3	−9.7	0.5

After generating the true-binary (B) network, drug targets lose 6.2 edges on average compared to their degree in the full PIN. At the same time, non-drug targets lose 4.9 edges. This difference is statistically significant (p-value = 9.3e-13) for the two first cases, therefore we conclude that removal of lower-confidence data preferentially decreases the degree of drug targets rather than non-drug targets.

ISSN: 1471-2105