Table 4.

Performance of various prediction models optimized for IMAC30 and CM10 chips for the BKVAN versus AR groups, BKVAN versus stable transplant groups, and AR versus stable transplant groupsa

Prediction ModelIMAC30 (%)CM10 (%)
BKVAN versus ARBKVAN versus ControlAR versus ControlBKVAN versus ARBKVAN versus ControlAR versus Control
SVM + all peaks
    ACE29.5623.1439.1441.6135.1640.56
    sensitivity61.0968.9758.9447.7754.9958.39
    specificity76.6482.1162.4765.4171.6360.36
SVM + 100 t test
    ACE16.8311.3317.4525.9430.9830.37
    sensitivity79.4083.7881.0865.3548.4268.03
    specificity86.4992.2384.0880.5476.0671.45
RF
    ACE27.5016.7622.9135.2834.6237.59
    sensitivity51.7572.7072.0143.2529.3257.02
    specificity88.8091.0482.4181.3491.2068.34
CART
    ACE29.0018.0029.0039.0020.0032.00
    sensitivity62.0090.0067.0042.0067.0056.00
    specificity76.0077.0075.0074.0090.0081.00
  • a The support vector machine (SVM) classifier was tested on all peaks (SVM + all peaks) and the top 100 differential peaks according to the P value of the t test (SVM + 100 t test). The other two models include Random Forest (RF) and Classification and Regression Tree (CART). The statistics listed include Achieved (average) Classification Error (ACE), sensitivity, and specificity.