Figure 1.

Insertion of annotation errors: Errors were randomly inserted into reference set annotations at a fixed error rate. The precision of reference set annotations for predicting the annotations of query sequences was determined, and the average precision at that error rate was recorded. This process was repeated 100 times for a given error rate value, after which the error rate was incremented. This process continued until data was obtained for artificially increased error rates of between 2% and 40%.