Skip to main content

Table 3 Coefficient of determination, r2, calculated for regression sets (higher values are better)

From: Transformer-CNN: Swiss knife for QSAR modeling and interpretation

DatasetDescriptor based methods2SMILES based (augm = 10)aTransformer-CNN, no augmTransformer-CNN, augm = 10CDDD descriptorsb
MP0.830.850.830.860.85
BP0.980.980.970.980.98
BCF0.850.850.71 ± 0.020.850.81
FreeSolv0.940.930.72 ± 0.020.910.93
LogS0.920.920.850.910.91
Lipo0.70.720.60.730.74
BACE0.730.720.660.760.75
DHFR0.62 ± 0.030.63 ± 0.030.46 ± 0.030.67 ± 0.030.61 ± 0.03
LEL0.19 ± 0.040.25 ± 0.030.2 ± 0.030.27 ± 0.040.23 ± 0.04
  1. We omitted the standard mean errors, which are 0.01 or less, for the reported values
  2. aResults from our previous study [22]. bBest performance calculated with CDDD descriptors obtained using autoencoder Sml2canSml from [27]