Skip to main content
. 2020 Oct 17;9(10):giaa110. doi: 10.1093/gigascience/giaa110

Table 2:

Cross-validation results

Antibiotic Lasso Cluster-lasso
bACC AUC Support Unitigs bACC AUC Support Unitigs
Amikacin 92.7 95.4 16 22 (4) 92.3 95.7 11 93 (36)
Aztreonam 76.7 81.9 31 45 (3) 76.9 82.3 28 425 (125)
Cefepime 74.0 80.4 53 65 (3) 73.6 79.8 34 385 (111)
Cefoxitin 82.4 88.7 134 155 (5) 82.2 88.6 171 1,052 (221)
Ceftazidime 91.6 95.8 51 69 (5) 90.7 95.3 43 863 (185)
Ciprofloxacin 95.6 98.6 25 27 (2) 95.5 98.6 35 422 (139)
Imipenem 93.1 93.6 10 10 (1) 92.7 93.4 7 241 (194)
Meropenem 91.7 94.0 8 8 (1) 91.4 93.5 3 164 (159)
Piperacillin/tazobactam 81.6 89.6 127 144 (4) 81.5 89.0 120 1,220 (226)
Tetracycline 83.0 88.5 181 198 (3) 82.9 87.7 109 640 (104)

This table summarizes the cross-validation results obtained by the lasso and cluster-lasso strategies for the 10 antibiotics, in terms of balanced accuracy (bACC), AUC, support size, overall number of unitigs involved, and maximal number of unitigs associated with a single pattern or cluster (in parentheses).