Automated methods of predicting the function of biological sequences using GO and BLAST

BMC Bioinformatics

Table 2 Accuracies of BLAST-based automated GO term predictors

Method	Precision	Recall	Harmonic Mean	Mean Term Assignments	N
Best BLAST	0.41	0.56	0.48	3.9	4710
Covering Graph
Ten highest scoring terms	0.20	0.48	0.28	7.7	4710
Terms with >0.1 normalised concordance	0.19	0.61	0.29	16.1	4710
Terms with >0.1 and <0.2 normalised concordance	0.09	0.10	0.09	2.7	4710
Terms with >0.1 and <0.9 normalised concordance	0.11	0.29	0.16	7.0	4710
Term Covariance Filter selection	0.24	0.35	0.28	3.4	4710
Term Distance Concordance
Ten highest scoring terms	0.33	0.63	0.43	6.3	4710
Term Covariance Filter selection	0.36	0.51	0.42	3.9	4710
Discriminant Function
Training set (post hoc)	0.70	0.59	0.64	2.4	2070
Test set (apriori)	0.61	0.51	0.55	2.0	10689

ISSN: 1471-2105