Skip to main content

Advertisement

Table 1 Model performance depends on datasets that are used in training

From: Identification of novel small molecule inhibitors for solute carrier SGLT1 using proteochemometric modeling

Model and validation Training Sensitivity Specificity PPV NPV MCC
QSAR (EV) PD + IH 0.76 0.86 0.42 0.96 0.48
Public PCM (CV) PD 0.01 ± 0.01 0.98 ± 0.00 0.03 ± 0.06 0.91 ± 0.01 − 0.03 ± 0.03
In-house PCM (CV) IH 0.69 ± 0.07 0.89 ± 0.02 0.38 ± 0.06 0.97 ± 0.01 0.45 ± 0.05
Combined PCM (CV) PD + IH 0.64 ± 0.06 0.93 ± 0.01 0.47 ± 0.07 0.96 ± 0.01 0.49 ± 0.05
  1. PD public data, IH in-house data, EV external validation on 30% of data, CV fivefold cross validation on 20% of the data per iteration