Skip to main content
. 2021 Aug 16;4(4):558–574. doi: 10.1089/crispr.2021.0021

Table 3.

Benchmarking Performance Results

Subtype Precision Recall F1 Repeats detected in test set Training examples
I-A 0.89 1.00 0.94 16 147
I-B 0.95 0.88 0.91 518 1,429
I-C 1.00 1.00 1.00 1,379 1,448
I-D 0.88 0.97 0.92 29 164
I-E 0.98 1.00 0.99 2,144 4,388
I-F 0.99 1.00 1.00 541 1,314
I-G 0.47 1.00 0.64 52 356
II-A 0.95 1.00 0.98 472 803
II-C 1.00 0.99 0.99 1,670 637
III-A 0.77 0.96 0.85 110 697
III-B 0.31 0.31 0.31 13 268
III-C 0.00 0.00 0.00 17 19
III-D 1.00 0.85 0.92 214 408
V-A 0.86 1.00 0.93 121 58
V-F 0.00 0.00 0.00 3 17
V-K 0.89 0.89 0.89 9 33
V-U2 0.00 0.00 0.00 8 6
V-U4 0.00 0.00 0.00 40 18
VI-A 0.00 0.00 0.00 23 20
VI-B 1.00 1.00 1.00 31 113

Performance metrics listed per subtype show high precision, recall, and F1 scores for subtypes with >30 training examples, with the exception of subtype III-B, which had 268 training examples but a low F1 score of 0.31.