Binding Site Prediction for Protein-Protein Interactions and Novel Motif Discovery using Re-occurring Polypeptide Sequences

Table 2 Binding site prediction error for two domain-oriented methods and PIPE-Sites (this work)

Species	Pairs originally in validation set	Method	Pairs with available predictions	Error (average DM)
Yeast	265	DPEA	32	0.718
Yeast	265	DOMINE	144	0.363
Yeast	265	PIPE-Sites (this work)	174	0.218
Human	423	DPEA	2	N/A *
Human	423	DOMINE	266	0.306
Human	423	PIPE-Sites (this work)	363	0.246

DPEA [12] is a standalone statistical inference method, while DOMINE [25] is a database combining several experimental and computational domain-domain interaction sources. We used a consensus approach to combine the 13 constituent data sources of DOMINE to arrive at a single binding site prediction. Error is expressed as average distance between the predicted site and lab-confirmed site, averaged over all pairs. * Due to the small sample size (only 2 available predictions), it was impossible to calculate a statistically significant average DM for DPEA predictions.

ISSN: 1471-2105