Table 3.

The classifier parameters are fixed by the choice from three scenarios responsible for determining the similarity between drugs, proteins, and their embedding vectors

Classifiers	KGE	KGE-ProtBERT	Molecular fingerprint and protein characteristics
ETC	n-estimators = trees, random-state = 1357	n-estimators = trees, random-state = 1357	n-estimators = trees, random-state = 1357
DT	random-state = 1357	random-state = 1357	random-state = 1357
MLP	solver = lbfgs, alpha = 1e−5, hidden-layer-sizes = (5, 2), random-state = 1	solver = lbfgs, alpha = 1e−5, hidden-layer-sizes = (240, 96), random-state = 1	solver = lbfgs, alpha = 1e−5, hidden-layer-sizes = (240, 96), random-state = 1
SGD	loss = log, penalty = l2, max-iter = 5	loss = log, penalty = l2, max-iter = 2	loss = log, penalty = l2, max-iter = 2
Gaussian-NB
Gradient Boosting	n-estimators = 100, learning-rate = 1.0,max-depth = 1, random-state = 0	n-estimators = 100, learning-rate = 1.0,max-depth = 2, random-state = 0	n-estimators = 100, learning-rate = 1.0,max-depth = 2, random-state = 0
Bagging Classifier	KNeighborsClassifier(), max-samples = 0.5, max-features = 0.5	KNeighborsClassifier(n-neighbors = 1),max-samples = 1, max-features = 1	KNeighborsClassifier(n-neighbors = 1),max-samples = 1, max-features = 1
K-Neighbors	n-neighbors = 7	n-neighbors = 2	n-neighbors = 2
RF	n-estimators = trees, n-jobs = 6, criterion = c, class-weight = balanced, random-state = 1357	n-estimators = trees, n-jobs = 6, criterion = c, class-weight = balanced, random-state = 1357	n-estimators = trees, n-jobs = 6, criterion = c, class-weight = balanced, random-state = 1357