Improving automatic GO annotation with semantic similarity

BMC Bioinformatics

Table 2 Effect of pruning and post-processing method on the performance of GrAPFI-GO and two other automatic annotation tools

Method	Post-processing cut-off	Precision	Recall	F1-score
GrAPFI	No-post-processing	0.165	0.108	0.107
	SS-max	0.573	0.115	0.175
	SS-5	0.445	0.380	0.376
	SS-5-MS-max/2	0.440	0.391	0.379
PANNZER	No-post-processing	0.547	0.942	0.668
	SS-max	0.637	0.225	0.301
	SS-5	0.634	0.515	0.536
	SS-5-MS-max/2	0.603	0.689	0.609
DeepGOPlus	No-post-processing	0.053	0.653	0.095
	SS-max	0.249	0.120	0.138
	SS-5	0.186	0.182	0.160
	SS-5-MS-max/2	0.167	0.233	0.1725

The bold numbers indicate which post-processing cut off achieved the maximum performance score for a particular performance metric and annotation tool
Average precision, recall and F1-score are computed for each method in four situations. No-post-processing: without post-processing and pruning; SS-max: pruned using highest SS as cut-off; SS-5: pruned using 5th highest SS as cut-off; SS-5-MS-max/2: pruned using 5th highest SS and (maximum MS)/2 as cut-offs

ISSN: 1471-2105