Table 2.
(A) Random Forest results for the classification model, (B) Identification of an independent test dataset from unknown samples
| Identified as | True | |||||
|---|---|---|---|---|---|---|
| Lag | Log | Stationary | Specificity (%) | SDM | ANOSIM | |
| (A) | ||||||
| Lag | 67 | 4 | 0 | 97.1 | 0.12 | 0.83 (lag, log)** |
| Log | 2 | 64 | 7 | 85.3 | 0.20 | 0.71 (log, stationary)** |
| Stationary | 0 | 7 | 63 | 90 | 0.15 | 0.89 (lag, stationary)** |
| Sensitivity (%) | 94.4 | 87.7 | 90 | 90.7 | ||
| Group | No. of correctly assigned cells | No. of total cells | Accuracy (%) | Total accuracy (%) |
|---|---|---|---|---|
| (B) | ||||
| Lag | 25 | 26 | 96.2 | |
| Log | 31 | 36 | 86.1 | 91.2 |
| Stationary | 27 | 29 | 93.1 | |
The standard deviation of the means (SDM) of SCRS per sample was calculated (range from 0.12 to 0.20). Low SDM numbers represent high reproducibility and high reliability of the dataset. An analysis of similarity (ANOSIM) was performed to compare distances of between-group cells and within-group cells at different growth states. ** Represents p < 0.01. Lag, lag phase, log, log phase, stationary, stationary phase