Skip to main content

Table 1 Current drug response prediction approaches

From: Quantitative Structure–Mutation–Activity Relationship Tests (QSMART) model for protein kinase inhibitor response prediction

Date Author Best model Compared models Cancer cell line features Drug response Validation Performance
2013/04/30 Menden et al. [10] ANN RF   \(\checkmark\) \(\checkmark\)   \(\checkmark\)   8-fold CV \(\hbox {R}^{2} = 0.72\)
2014/01/01 Jang et al. [11] GLM RF, SVM, PCA, PLS \(\checkmark\) \(\checkmark\) \(\checkmark\) CLS \(\checkmark\) \(\checkmark\) 5-fold CV r = \(\sim\) 0.5
2014/03/03 Geeleher et al. [12] GLM   \(\checkmark\)     \(\checkmark\)   LOOCV AUC = 0.81
2015/06/30 Dong et al. [13] SVM   \(\checkmark\)      \(\checkmark\) 10-fold CV Accuracy = \(\sim\) 0.8
2015/09/29 Zhang et al. [14] Network EN \(\checkmark\)     \(\checkmark\) \(\checkmark\) LOOCV r = 0.6
2016/03/31 Gupta et al. [15] SVM   \(\checkmark\) \(\checkmark\) \(\checkmark\)    \(\checkmark\) LOOCV r = 0.78
2016/09/01 Ammad-Ud-Din et al. [16] Kernel GLM     PWY \(\checkmark\)   5-fold CV \(\rho\) = \(\sim\) 0.22
2016/12/28 Nguyen et al. [17] MANOVA RF \(\checkmark\)     \(\checkmark\)   10-fold CV MCC = 0.18
2017/01/09 Stanfield et al. [18] Network Kernel   \(\checkmark\)   PPI \(\checkmark\) \(\checkmark\) LOOCV AUC = 0.881
2017/07/15 Ammad-Ud-Din et al. [19] GLM RF, SVM, PLS, SGL \(\checkmark\)    PWY \(\checkmark\)   LOOCV \(\rho\) = 0.375
2017/08/28 Geeleher et al. [20] Ridge   \(\checkmark\)     \(\checkmark\)   10-fold CV \(\rho\) = 0.48
2017/09/12 Rahman et al. [21] RF   \(\checkmark\)     \(\checkmark\) \(\checkmark\) 3-fold CV AUC = \(\sim\) 0.3
2018/02/01 Ding et al. [22] DNN EN, SVM \(\checkmark\) \(\checkmark\) \(\checkmark\)   \(\checkmark\) \(\checkmark\) 25-fold CV AUC = 0.87
2018/06/11 Chang et al. [23] CNN RF, SVM     SNP \(\checkmark\)   5% leave-out \(\hbox {R}^{2}\) = 0.843
2018/07/01 Cichonska et al. [24] Kernel   \(\checkmark\)   \(\checkmark\) SNP, MET \(\checkmark\)   10-fold CV r = 0.858
2018/08/15 He et al. [25] Kernel RF, EN, Ridge \(\checkmark\)     \(\checkmark\)   3-fold CV Precision = \(\sim\) 0.35
2018/09/14 Juan-Blanco et al. [26] Network   \(\checkmark\) \(\checkmark\)     \(\checkmark\) LOOCV AUC = \(\sim\) 0.72
2018/09/14 Le and Pham [27] Network Kernel \(\checkmark\) \(\checkmark\)    \(\checkmark\) \(\checkmark\) 5-fold CV r = 0.804
2018/12/07 Liu et al. [28] Network   \(\checkmark\)     \(\checkmark\) \(\checkmark\) 10-fold CV r = 0.73
2019/01/22 Wei et al. [29] Network   \(\checkmark\)     \(\checkmark\) \(\checkmark\) LOOCV r = 0.63
2019/01/31 Wang et al. [30] EN   \(\checkmark\)    PWY   \(\checkmark\) 10-fold CV MSE = \(\sim\) 2.8
2019/01/31 Chiu et al. [31] DNN SVM, PCA, LR \(\checkmark\) \(\checkmark\)    \(\checkmark\)   10% leave-out r = \(\sim\) 0.86
2019/02/27 Li et al. [32] Mixture RF, GLM \(\checkmark\)      \(\checkmark\) 20% leave-out r = 0.882
2019/05/01 Yang et al. [33] Network + SVM Kernel   \(\checkmark\) \(\checkmark\) PPI, MET \(\checkmark\)   5-fold CV AUC = 0.788
2019/07/11 Lind and Anderson [34] RF ANN, SVM   \(\checkmark\)    \(\checkmark\)   5-fold CV r = 0.86
2019/07/29 Liu et al. [35] CNN ANN   \(\checkmark\) \(\checkmark\)   \(\checkmark\)   10% leave-out \(\hbox {R}^{2}\) = 0.826
2019/10/31 Manica et al. [36] MCA + CNN RF, SVM \(\checkmark\)   \(\checkmark\) PPI \(\checkmark\)   5-fold CV \(\underline{\textit{R}^{\textit{2}}\,=\,0.86}\)
2019/11/04 Oskooei et al. [37] Network RF, LR \(\checkmark\)    PPI \(\checkmark\)   30-fold CV r = \(\sim\) 0.9
  1. The best performing method is highlighted in underlined
  2. ANN, artificial neural network; AUC, area under the ROC curve; CCLE, Cancer Cell Line Encyclopedia; CLS, cancer classification; CNN, convolutional neural network; CNV, copy number variation; CV, cross-validation; EN, elastic net; EXP, gene expression; GDSC, Genomics of Drug Sensitivity in Cancer; GLM, generalized linear model, including ridge, elastic net, and lasso regression; DNN, deep neural networks; LOOCV, leave-one-out cross-validation; LR, linear regression; MCA, multiscale convolutional attentive; MCC, Matthews correlation coefficient; MET, methylation; MSE, mean squared error; MUT, gene-level mutation (i.e. whether the gene is mutated or not); PCA, principal component analysis; PLS, partial least squares; PPI, protein–protein interaction; PWY, pathway; r, Pearson correlation coefficient; \(\hbox {R}^{2}\), coefficient of determination; RF, random forests; \(\rho\), Spearman’s rank correlation coefficient; RNN, recurrent neural network; SGL, sparse group lasso; SNP, single nucleotide polymorphism; SVM, support vector machine