Figure 1.

Predicting the number of domains. (A) We compared the success in predicting the number of domains to random predictions. Our method was significantly better than random for all multi-domain proteins. (B) For 69% of all and for 38% of all multi-domain proteins the number of domains predicted by CHOPnet and observed by CATH were identical. Overall, CHOPnet appeared to slightly over-predict the number of domains (solid bars on right higher than those on left). However, the data for multi-domain proteins revealed that it actually under-predicted significantly, in particular the number was under-predicted more often than it was predicted correctly (stippled bar at ‘−1’ higher than that at ‘0’).