TABLE 2

Summary of model performance of in silico VD,ss prediction methodologies for Lombardo intravenous dosing drug set (n = 1352 drugs) divided into two subsets: 1) ATOM in silico set (>940 compounds) and 2) ATOM experimental set (n > 280 compounds)

Method DescriptionInput ParametersnWithin 2-FoldWithin 3-FoldWithin 10-Foldr2AAFE
%
ADMET mechanisticLog D, fup, BPR (predicted using ADMET Predictor models)9564765900.252.8
2875171920.352.4
ATOM mechanisticLog D, fup, BPR (predicted using ATOM ML models)9364562870.233.1
2855066920.382.7
Allometry (rat and dog)Rat VD,ss, dog VD,ss (predicted using ATOM ML models)9564566930.282.7
2834569960.372.5
Allometry (rat)Rat VD,ss, rat fup, fup (human) (predicted using ATOM ML models)9563047790.024.4
2832340760.04.9
Allometry (dog)Dog VD,ss, dog fup, fup (human) (predicted using ATOM ML models)9562843750.054.9
2832338770.124.7
Direct ML model for human VD,ssSMILES/MOE descriptors9563655880.143.3
2855875980.522.2