Skip to main content
. 2025 Jun 26;16:1605722. doi: 10.3389/fphar.2025.1605722

TABLE 1.

Performance Metrics of best performing models developed for NS3 and NS5 Protein using various MLTs and Feature Selection method on TT and IV Datasets.

Algorithm Feature selection Model parameters Dataset MAE MSE RMSE R 2 PCC
NS3
SVM SVR gamma:0.0001 C:100 T 1092 0.191 0.065 0.255 0.733 0.857
V 121 0.195 0.074 0.272 0.756 0.870
Perceptron gamma:0.001 C:10 T1092 0.239 0.099 0.315 0.593 0.771
V121 0.212 0.087 0.295 0.713 0.844
DT gamma:0.005 C:1 T1092 0.255 0.118 0.343 0.517 0.719
V121 0.215 0.097 0.311 0.680 0.833
RF SVR n:400 T1092 0.264 0.124 0.352 0.492 0.703
V121 0.213 0.093 0.305 0.693 0.851
Perceptron n:500 T1092 0.275 0.130 0.361 0.467 0.684
V121 0.225 0.102 0.319 0.663 0.832
DT n:400 T1092 0.266 0.124 0.350 0.475 0.705
V121 0.218 0.103 0.321 0.66 0.825
kNN SVR k:3 T1092 0.251 0.119 0.346 0.511 0.725
V121 0.250 0.131 0.363 0.565 0.756
Perceptron k:7 T1092 0.267 0.127 0.356 0.480 0.695
V121 0.227 0.105 0.325 0.651 0.811
DT k:5 T1092 0.270 0.131 0.362 0.464 0.690
V121 0.233 0.109 0.330 0.639 0.799
ANN SVR solver:lbfgs
activation: identity		 T 1092 0.192 0.064 0.253 0.738 0.862
V 121 0.190 0.070 0.265 0.781 0.894
Perceptron activation:logistic T1092 0.249 0.108 0.329 0.556 0.751
V121 0.216 0.086 0.293 0.716 0.844
DT activation: logistic T1092 0.275 0.135 0.365 0.426 0.682
V121 0.241 0.111 0.332 0.634 0.798
XGBoost SVR n_estimators = 300, max_depth = 3, learning_rate = 0.141 T1092 0.249 0.111 0.334 0.544 0.738
V121 0.222 0.087 0.296 0.710 0.849
Perceptron n_estimators = 300, max_depth = 7, learning_rate = 0.058 T1092 0.272 0.132 0.362 0.448 0.678
V121 0.212 0.089 0.298 0.705 0.842
DT n_estimators = 200, max_depth = 3, learning_rate = 0.104 T1092 0.259 0.119 0.344 0.514 0.718
V121 0.225 0.096 0.31 0.681 0.830
NS5
SVM SVR gamma:0.0001 C:400 T 140 0.135 0.049 0.197 0.954 0.982
V 15 0.138 0.044 0.210 0.94 0.970
Perceptron gamma:0.0005 C:400 T140 0.222 0.105 0.310 0.884 0.953
V15 0.24 0.137 0.370 0.814 0.904
DT gamma:0.005 C:10 T140 0.429 0.399 0.591 0.632 0.802
V15 0.420 0.446 0.668 0.395 0.713
RF SVR n:400 depth: 12 T140 0.399 0.324 0.544 0.659 0.840
V15 0.340 0.211 0.46 0.713 0.863
Perceptron n:300 T140 0.360 0.294 0.513 0.680 0.852
V15 0.288 0.180 0.425 0.755 0.873
DT n:200 depth: None leaf:1 T140 0.424 0.388 0.601 0.560 0.799
V15 0.367 0.308 0.555 0.582 0.771
kNN SVR k:3 T140 0.343 0.235 0.468 0.727 0.889
V15 0.292 0.144 0.380 0.804 0.901
Perceptron k:3 T140 0.335 0.235 0.468 0.753 0.895
V15 0.360 0.232 0.481 0.686 0.833
DT k:3 T140 0.508 0.499 0.687 0.446 0.739
V15 0.563 0.609 0.781 0.173 0.663
ANN SVR solver: lbfgs activation: identity learning: invscaling T 140 0.159 0.073 0.271 0.928 0.964
V 15 0.160 0.048 0.219 0.935 0.977
Perceptron solver: lbfgs activation: logistic learning: adaptive T140 0.255 0.119 0.345 0.884 0.942
V15 0.337 0.238 0.488 0.710 0.854
DT solver: lbfgs activation: tanh learning: adaptive T140 0.532 0.505 0.711 0.508 0.762
V15 0.507 0.592 0.769 0.197 0.708
XGBoost SVR n_estimators = 300, max_depth = 3, learning_rate = 0.078 T140 0.334 0.22 0.444 0.766 0.889
V15 0.388 0.274 0.523 0.628 0.818
Perceptron n_estimators = 300, max_depth = 3, learning_rate = 0.143 T140 0.335 0.243 0.458 0.678 0.846
V15 0.313 0.200 0.447 0.729 0.870
DT n_estimators = 150, max_depth = 5, learning_rate = 0.038 T140 0.406 0.326 0.548 0.609 0.812
V15 0.358 0.207 0.455 0.719 0.854

SVR, Support vector regression; DT, Decision tree; TT, Training or testing dataset; IV, Independent validation dataset; PCC, Pearson’s correlation coefficient, R 2 - coefficient of determination, MAE, mean absolute error; MSE, mean squared error; RMSE, root mean squared error.

Bold values indicate the best-performing model(s) based on Pearson correlation coefficient (PCC) and coefficient of determination (R2) for each dataset.