Data Supplement
- Supplemental Data -
Supplemental Table S1 - Machine learning (ML) models used to predict input parameters for various in silico VD,ss prediction methodologies
Supplemental Table S2 - Experimental measurement of physicochemical properties and observed VD,ss values (Lombardo et al., 2018) for 254 compounds
Supplemental Table S3 - Experimental determination of adipocyte and myocyte cell partitioning of 189 clinical compounds (Lombardo et al., 2018).
Supplemental Figure S1 - UMAP projection of datasets (a) ATOM in silico and (b)experimental set based on ECFP Tanimoto distances.
Supplemental Figure S2 - Scatter/KDE plots of observed versus predicted using various silico VD,ss prediction methodologies listed in Table 2 (a) ATOM in silico set (>940 compounds) and (b) ATOM experimental set (n>280 compounds). Crosslines indicate 2, 3 and 10-fold limits.
Supplemental Figure S3 - Measured ChromLogD compared to predicted LogD values by (a) ADMET Predictor (b) ATOM ML model.
Supplemental Figure S4 - Scatter/KDE plots of mechanistic VD,ss predictions using various combinations of experimental data (fup, BPR and LogD) as input parameters.
Supplemental Figure S5 - Scatter/KDE plots of mechanistic VD,ss predictions using various combinations of experimental data (Kp fat and Kp muscle) as input parameters.
Supplemental Figure S6 - Scatter plot of mechanistic VD,ss predictions color coded by ionization using various combinations of experimental data (fup, BPR and LogD) as input parameters.
Supplemental Figure S7 - Scatter plots of predicted in silico vs observed VD,ss using (a) Allometry (Rat and Dog) and (b) ADMET Mechanistic models. Crosslines indicate 2, 3 and 10-fold limits.