Table 1 The cutting-edge ML-based Kglu prediction methods
Tool | features extraction/selection | balanced/classification algorithm | Performance parameters | AUC (%) | Acc (%) |
|---|---|---|---|---|---|
GlutPred [7] | AAF + BE + CKSAAP mRMR + IFS | Bias SVM | ten-fold cross-validation | 78.06% | 74.90% |
iGlu_Lys [9] | PSPM | SVM | ten-fold cross-validation | 89.44% | 88.38% |
MDD_Glutar [10] | ACC | SVM | five-fold cross-validation | 63.74% | 61.60% |
BiPepGlut [11] | bi-peptide-based PSSM | Extra-Trees | ten-fold cross-validation | — | 74.58% |
PUL-GLU [8] | AAF + BE + CKSAAP | Positive-unlabeled Learning/ SVM | ten-fold cross-validation | 85.30% | 81.50% |
RFGlutarySite [12] | PseAAC + CT + SE + RE + IG + CTD + AAC + DC + TC + Autocorrelation、BE + AAindex + AAF + CKSAAP/Xgboost | Random Forest | ten-fold cross-validation | 81.00% | 72.30% |
DEXGB_Glu [14] | AAindex, + ASA + SS + PSSM、RC、AC | Borderline-SMOTE/Xgboost | ten-fold cross-validation | — | 87.09% |
iGlu_AdaBoost [13] | 188D + CKSAAP + EAAC | SMOTE-Tomek /Adaboost | ten-fold cross-validation | 89.00% | 79.98% |
iGluK-Deep [15] | PseAAC | FCN | — | — | 94.30% |
ProtTrans-Glutar [16] | CTDD + EAAC + ProT5-XL-UniRef50 | RUS/XGBoost | ten-fold cross-validation | 70.75% | 65.67% |
DeepDN_iGlu [17] | BE | focal loss/DenseNet | ten-fold cross-validation | 77.25% | 66.00% |
Deepro-Glu [18] | BE + DDE + BLOSUM62 + AAindex + ProtBert | Attention + MLP | ten-fold cross-validation | 98.80% | 96.30% |