Table 1. Over and under predictions of EPPIC and PISA in the PDB1 and PiQSi datasets.

	PDB1 dataset		PiQSi
	EPPIC	PISA	EPPIC	PISA
Correct	85.3%	83.5%	72.8%	79.1%
Over-predicted	4.2%	13.2%	3.0%	12.3%
Under-predicted	10.5%	3.4%	24.2%	8.6%

Over and under predictions of EPPIC and PISA in the PDB1 assemblies dataset, introduced in this article, and the PiQSi dataset. Over predictions correspond to the upper-left half of the confusion matrix in Fig 3, whilst under predictions correspond to the lower-right half.