Table 10 The comparative study between recent approaches and the proposed model using the same dataset in this study.
From: Harnessing machine learning to find synergistic combinations for FDA-approved cancer drugs
Author | Method | Results | Comment |
|---|---|---|---|
Li et al.69 | SNRMPACDC: Siamese CNN, Random Matrix Projection, MLP | RMSE: 15.01, Pearson correlation: 0.75 (regression) | Good regression & classification (AUC 0.91, AUPR 0.62) |
Huang et al.70 | Kaplan–Meier, Cox Regression, Nomogram | AUC > 0.800 for 3-, 5-, and 10-year intervals | Well-calibrated nomogram for clinical use |
Zhang et al.71 | Sequential model with various feature encodings | M13-M20 performance metrics documented | Comprehensive feature analysis, confidence intervals included |
Kuru et al.72 | MatchMaker: Deep learning with drug structure & gene expression | DrugComb: 15% correlation, 33% lower MSE | Deep learning with large dataset and good performance |
Zagidullin et al.73 | Drug and cell line categorization in DrugComb | 33.3% of drugs lacked documented mechanisms | Highlighted data limitations in DrugComb |
Tang and Gottlieb74 | Biologically motivated deep learning for pathway features | MSE: 70.6 ± 6.4, topologically interacting pathways for synergy | Considered pathway interactions for improved prediction |
El Khili et al., 2023 75 | MARSY: Deep learning multitask model for synergy prediction | RMSE: 9.06 (± 0.45) for drug-pair combinations | Efficient prediction for large datasets with multitask learning |
Proposed model | Classification (NB, RF, KNN, and LR) | Average Accuracy: 89% | After applying data preprocessing |
Regression (Linear, Ridge, and RF regressors) | Average MAE: 0.0984 R2: 0.5290 MSE:0.015 | Prediction methods were employed to forecast the CSS score of drug combinations |
