The functional annotation of Arabidopsis protein sequences was performed by BLAST queries against a reference set of experimentally verified enzymes.
For each Arabidopsis sequence, the enzymatic activity of the top BLAST hit (or hits if they had equivalent E-values) was assigned to the protein if its
E-value fell below a specific E-value threshold established for the corresponding enzymatic activity. Note: The annotation thresholds were
established by doing a self BLAST of the reference enzyme dataset. For each enzymatic activity represented by multiple proteins, the mean E-value of all
the correct hits generated by the self BLAST was selected as the cut-off. All of these means were averaged and used as the cut-off for assigning
annotations for any enzymatic activities that were represented by a single protein in the reference dataset.
Chi A, Rhee S