. 2023 Mar 20;19(14):4701–4710. doi: 10.1021/acs.jctc.2c01303

Table 2. CNS MPO Average Score Comparison at the Same Epochs as in Table 1 ^a.

	Metric^b
Model (learning rate)	Epoch	Calc. CNS MPO	Change from original	Predicted CNS MPO	Change from original
GGNN (l × l0^–4)	72	0.964	–0.011	5.090	–0.170
GGNN (l × l0^–5)	66	0.963	–0.012	5.195	–0.075
MNN (l × l0^–4)	90	0.973	–0.002	5.337	+0.077
MNN (l × l0^–5)	84	0.972	–0.003	5.330	+0.070

The calculated CNS MPO score ranges between 0 and 1, 1 implying a very high probability of BBB penetrance, while the predicted CNS MPO score ranges between 0 and 6.

Metrics are reported for the optimally performing epoch. Each metric has an associated “Change from original” column, which refers to the mean change between the generated population and training set.

Table 2. CNS MPO Average Score Comparison at the Same Epochs as in Table 1a.

Table 2. CNS MPO Average Score Comparison at the Same Epochs as in Table 1 ^a.