Table 3 Evaluation criteria for GO and protein predictions.

High

The GO term assignment was correct or close to what a curator would choose, given the evidence text.

The protein mentioned in the evidence text correctly represented the associated UniProt accession (correct species).

General

The GO term assignment was in the correct lineage, given the evidence text, but was too high level (parent of the correct GO term) e.g. biological_process or too specific.

The evidence text did not support annotation to the associated UniProt accession but was generally correct for the protein family or orthologs (non-human species).

Low

The evidence text did not support the GO term assignment. Note: The GO term may have been correct for the protein but the evidence text did not support it.

The evidence text did not mention the correct protein (e.g. for Rev7 protein (ligand) incorrect evidence text referred to 'Rev7 receptor') or protein family.