|
|
 |
The annotation of sequences is performed according to the "best" BLAST score, i.e. for each sequence from the test set
we select the annotated sequence from the training set that has maximum BLAST score (i.e., highest similarity). After this we assign all FunCat categories of the training set
sequence to the sequence from the test set. The confidence value is 1/SCORE, where SCORE is the "best" BLAST score.
Thus this annotation corresponds to the use of k-nearest neighbor classifier with K=1. |
|
Performance of the method
genome |
total genes |
coverage of manually annotated genes |
annotations of new genes |
sensitivity |
specificity |
|
Helicobacter_pylori |
1576 |
78.7 (685 out of 870) |
180 |
73.9 |
51.6 |
Bacillus_subtilis |
4112 |
74.7 (2108 out of 2823) |
214 |
79.7 |
82.8 |
Listeria_monocytogenes |
2846 |
86.9 (1693 out of 1948) |
157 |
89.1 |
85.8 |
Listeria_innocua |
2968 |
86.9 (1697 out of 1953) |
170 |
89.1 |
85.6 |
| |
 |
BFAB
