Fig. 1

Inductive conformal predictor (ICP) and the aggregated conformal prediction methods used in this study. a Aggregated conformal prediction (ACP) and ICP (box with purple edge): The dataset is split into a training set and a test set. The training set is further split into a proper training set to train the model and a calibration set. The predictions made for the test set compounds are used to calculate nonconformity scores (nc) and compared to nonconformity scores in the calibration set to calculate p-values and generate prediction sets. In ACP, multiple models are trained and calibrated with randomly selected proper training and calibration sets, and p-values from these are averaged. b Synergy conformal prediction (SCP): In order to ensure a uniform distribution of p-values, SCP averages the nonconformity scores instead. Multiple models are trained on (subsets of) the proper training set and with each model predictions are made for the test set and for a fixed calibration set