How can we determine the reliability of the predicted targets of BindingDB, when different methods of accessing them yield different results for the same SMILES? For example, using the ‘Find My Compound’s Targets’ option, we obtain 4 targets for Aspirin, while using the ‘Batchquery’ file from the same option, we obtain 12 targets, and using the API (getTargetByCompound), we obtain 6 targets.