Table 8. Straw and baselines models experiment on BiComp-DTA for the Davis dataset.

Experiments		CI(std)	MSE	CI loss	MSE loss
Straw models	S1	0.506 (0.004)	0.794	44%	0.557
	S2	0.501 (0.015)	0.811	44%	0.574
	S3	0.494 (0.003)	1.289	44%	1.052
Baselines	KNN	0.845 (0.002)	0.364	6%	0.127
	RF	0.868 (0.001)	0.317	3%	0.080
	FC	0.855 (0.003)	2.790	5%	2.553
BiComp-DTA		0.904 (0.001)	0.237	-	-

S1, S2, and S3 denote different models, for which, the model training and testing are performed on the shuffled binding affinity values, the model training is performed on the shuffled binding values, and the model testing is performed on the shuffled binding values, respectively. KNN, RF, and FC denote k-nearest neighbors algorithm, random forests, and fully-connected neural network, respectively