Fig 1.
Overall flowchart (A) and the representative steps of procedures of dimension reduction and cross validation steps (B).
Table 1.
Top 10 peaks with least sum of 95 ranks produced each time one different spectrum omitted and their percentage of presence in each group.
Fig 2.
Performance of five machine learning algorithms with leave-one-out cross validation using 46 CRKP and 49 CSKP mass spectra, in terms of accuracy, sensitivity and specificity for differentiate CRKP from CSKP.
Accuracy (Accu), sensitivity (Sen) and specificity (Spe) are used to evaluate prediction systems. Panel A, Boxplots (25th to 75th percentiles, Min to max) of accuracy, sensitivity, and specificity for differentiation of CRKP from CSKP when 1–100 peaks selected. In Panel B, Values of accuracy, sensitivity and specificity with number of ranked peaks (k = 1–100) with increasing p-value (X axis) selected for machine learning algorithms. k: number of ranked peaks selected with increasing p-value for classification by using all five machine learning algorithms.
Table 2.
Performance of five machine learning algorithms with L1O cross validation using 46 CRKP and 49 CSKP mass spectra, in terms of accuracy, sensitivity and specificity for differentiate CRKP from CSKP.
Fig 3.
Random amplified polymorph in DNA fingerprinting (RAPD) types of 46 carbapenem-resistant K. pneumoniae generated by arbitrarily primed PCR.
Lanes S, standard strain included in every experiment as a control. Results for the study strains were shown in the order of the date of isolation. Lane M shows the 1-kb DNA ladder.