Table 2.

Comparative performance for predicting DNA- and peptide-binding positions within domains

Ligand	Method	AUPRC	AUC	Mean domain AUPRC	Mean domain AUC
DNA	dSPRINT	0.5450	0.8853	0.3866	0.7618
	Agg. Dr PIP	0.2697	0.8219	0.2696	0.7843
	Agg. DRNApred	0.1262	0.7176	0.2011	0.7175
peptide	dSPRINT	0.1522	0.6836	0.2345	0.6331
	Agg. SCRIBER	0.0409	0.5256	0.0867	0.4265
	Agg. SSPIDER	0.0545	0.4668	0.1255	0.4606

dSPRINT’s performance on independent benchmark datasets of 727 DNA-binding domain positions and 1,522 peptide-binding domain positions as compared to baseline methods that aggregate, by domain position, per-amino acid DNA or peptide-binding predictions made by state-of-the-art methods for predicting interaction sites within protein sequences. For all methods, AUPRC and AUC is computed based on positions across all domains together, and the mean per-domain AUPRC and AUC is computed based on considering positions in each domain separately and then averaging. For three out of four measures, dSPRINT outperforms aggregating DNA-binding scores by either DRNApred or Dr PIP, and for all four measures dSPRINT outperforms aggregating peptide-binding scores by either SCRIBER or SPPIDER.