Table 1. Domain number prediction accuracy (set_CATH)^a.

Observation (CATH assignment)	Prediction
	1	2	3
1	73 ± 4	25 ± 5	2 ± 2
2	54 ± 8	41 ± 7	5 ± 4
3	50 ± 12	29 ± 6	21 ± 12

^aPercentage of proteins observed with N (1–3) domains (rows) and predicted with M domains (column). The diagonal highlights correct predictions in bold. For example, 41% of the two-domain proteins were correctly predicted as such, for 54% the domain boundary was missed incorrectly predicting single domains and 5% were over-predicted to contain three instead of two domains.