An evaluation of GO annotation retrieval for BioCreAtIvE and GOA

BMC Bioinformatics

Table 6 Comparison of BioCreAtIvE test set manual annotations with electronic GO annotation predictions.

	InterPro2GO	SPKW2GO	EC2GO
Total IEA annotations	635	385	27
Exact term	151 (0.24)	62 (0.16)	18(0.67)	Correct
Same lineage > granularity	24 (0.04)	10 (0.03)	3 (0.11)	Potentially Incorrect/Correct
Same lineage < granularity	273 (0.43)	170 (0.44)	1 (0.04)	Correct
Total same lineage	297 (0.47)	180 (0.47)	4 (0.15)	Potentially Incorrect/Correct
New lineage	187 (0.29)	143 (0.37)	5 (0.19)	Potentially Incorrect/Correct
Total potential incorrect	211 (0.33)	153 (0.40)	8 (0.30)
Total minimal correct	424 (0.67)	232 (0.60)	19 (0.70)
Precision	0.67–1.00	0.60–1.00	0.70–1.00

Where the GO evidence code IEA is 'Inferred from Electronic Annotation' [27]. 'Same lineage > granularity' means where the electronic mapping (InterPro2GO, EC2GO or SPKW2GO) predicted a GO term that was in the same lineage/branch as the manually curated GO term but represented a more granular/parent term. 'Total potential incorrect' annotations = 'Same lineage >granularity' + 'New lineage'. 'Total minimal correct' annotations = 'Exact term' + 'Same lineage < granularity'. Percentage calculations are represented in parentheses.

ISSN: 1471-2105