Table 2.

AI in the biomarker discovery for PC diagnosis

Algorithm	Sample size	Data source	Result	Reference
RF	489 samples	RNA seq	AUC 0.945	Mahawan et al. [40]
SVM	107 cases	MS	Accuracy 97.4% AUC 0.997 Sensitivity 100% Specificity 95.0%	Iwano et al. [41]
1D CNN + LSTM	590 cases	Proteomics	Accuracy 97% AUC 0.98	Karar et al. [42]
GLM	208 cases	Proteomics	AUC 0.95 Sensitivity 84% Specificity 95%	Athanasiou et al. [43]
GBM	1023 cases	ELISA	AUC 0.919 PDAC vs. non-PDAC AUC 0.925 PDAC vs. healthy controls	Firpo et al. [44]
SVM	1504 cases	mRNA	AUC 0.985 resecable PC AUC 0.967 all PDAC stages	Lee et al. [45]
SVM	501 cases	RNAseq	AUC of 0.936 Sensitivity of 93.68% Specificity of 91.57%	Yu et al. [46]
RF	100 cases	SMS	AUC 0.977 Sensitivity 0.94 Specificity 0.94	Chen et al. [47]
OPLS-DA	830 samples	Lipidomic MS	Accuracy 94.18% AUC 0.983 Sensitivity 95.97% Specificity 90.46%	Worlab et al. [48]

1D CNN, one-dimensional convolutional neural network, AUC, Area under the curve, DL, Deep learning, ELISA, Enzyme-linked immunosorbent assay, GBM, Generalized boosted regression model, GLM, Generalized linear model, LSTM, Long short-term memory, MS, Mass spectrometry, OPLS-DA, Orthogonal partial least squares discriminant analysis, RF, Random forest, SMS, Shotgun metagenomic sequencing, SVM, Support vector machine